The reason people aren’t freaking out is because most people are using heavily subsidized subscriptions.
I tried the cheapest provider on openrouter and burned through $50 in a few days. Quality was ok, seems slightly above Luna quality perhaps? But that $50 is 1/4 of my codex subscription where I could have burned that many tokens or more using Astra within my weekly reset.
This won’t last forever but as long as the frontier labs are subsidizing this heavily the open models won’t matter.
Yes. You have to find the provider with pricing that suits your usage.
I am having 98% my input in cache, so using Coralbricks makes sense due to them giving cache reads for free — you only pay for writes. I spend maybe 5-10 dollars a day and my agents basically work day and night implementing things for me.
If your tasks are write-heavy, find a provider with cheaper output.
If you build a customer-facing app, pay a bit extra for 400+ tok/s e.g. on Lithos.
But aren't you developing bad habits and learning patterns that won't work long term? Or do you think things will get cheap enough that you will be able to keep going with your current patterns post-subsidies?
I expect by that point we'll have local models that can do a decent job, I would guess give it a decade and we'll be running custom accelerators that are smarter than current frontier models.
In the same way that only supercomputers used to have multiple processors and caches but it's now standard.
> But aren't you developing bad habits and learning patterns that won't work long term?
2 reasons - there's an advantage now, use it. 2nd the frontier providers, this is the "early cheap days" like when uber was initially cheap to compete vs standard cabs. they want you to become hooked and boy are we hooked.
I am curious how you managed to spend that much on Deepseek via OpenRouter. I loaded $100 back in July while using v4-flash or whatever the cheap good model was at the time, and have upgraded as the new ones came out from Deepseek. I still have $16 and some of that spend also goes towards the AI usage from my customers (the context they need to load in is quite large too).
And I am using the Claude Code harness with DS as the endpoint. And I use it ~5-8hrs a day to do my coding.
Likely the user doesn't know what they're doing or has extermely bad workflows. They're prob not managing their cache, and dont use compaction.. Letting context get to 500k and invalidating their cache every 10 tool calls because they have no providor fallback settings.
I was running deepseek v4.1 pretty much non stop during work hours, with heavy tool/mcp usage and finding it very difficult to spend more than $75 in a month.
Also the cheapest providers on Openroutrr can often have terrible cache hit %, short TTLs resulting in their effective price being much more expensive than people realize. 75% cache pretty much destroys any savings from a super cheap token perspective.
When 98.5% of my requests are cache hits (according to Pi for the last week), the cache miss price isn’t that important to me, and $0.003-0.006 per 1M input tokens is shockingly cheap.
It’s also the major difference between using DeepSeek directly vs other providers also serving it, though I have not looked lately: it’s possible other providers have matched its cache hit pricing better?
I've heard that certain inference providers may have different quality of caching implementations, so even if the listed numbers are as you say, the practical cache hit % you get might be significantly different/incur significantly different costs.
No, it’s outrageously profitable above x% without stealing any prompts. Provider economics still pretty good. Acquiring hardware is the current limiter.
VRAM aint cheap, Sam Altman ruined the cost of memory, Nvidia doesnt make enough consumer GPUs letting the market go insane over them, I still have friends on 1070s or 1070 TIs because GPUs have been severely overpriced for too long. I remember when a gaming PC was only $1000.
Even so why would anyone not sleep on a model they cannot run?
Seriously, if a single politician stepped forward and said "i'll bring down ram prices" they could then shoot a puppy and call me a slur and I'd still go out and doorknock for them.
Look at the financials of these companies, they're all making obscene margins and do they plan to increase production? No. Micron is doing a stock buy back to pump the price of their share.
There's sanctions, tarrifs, and a DOJ who doesn't give a shit. Until we can fix that the insanity will continue. Phones will be unaffordable. Laptops will be obscene. Gaming consoles will be thousands of dollars. Desktops will be dead.
If you're waiting for some David Ricardo equation to happen, tough cookies, it's not coming.
The market is legally locked down and we're in hostage pricing mode.
And what's the story? You can't afford electronics because we're using it to build robots to take your job? I mean ...
Micron has 3 brand new fabs currently under construction, 2 Boise, 1 in New York as the first of 4 planned for a campus.
Plus expanding other existing facilities.
These things take ~3-5 years from breaking ground to full production. You'd have had to anticipate the current demand years before it happened in order to be bringing production on-line before 2030 or so.
Samsung and HK Hynix also have fabs under construction and planned.
CXMT started 11 years ago and only now is reaching any real volume. If they decided a year ago to react to the current demand cycle they'd be 6-7 years out.
Not much you can really do to wish for more fabrication to exist on any timeline not measured in fractional decades.
Could they do more and react quicker? Probably, but everything I've read on the subject seems to point to 3 years is absolute bare minimum if you happen to have a shovel ready project with the land bought, local permitting completed, infrastructure extended to the site, and a skilled workforce already in place. They could suspend buy-backs/dividends today and dump it all into building production and there would be no material impact until around 2030.
> The Micron CEO just recently said this is the exact plan
CEO simply stated the demand pressure will not go away through 2027, and supply will not increase until around 2028 when currently under construction fabs start shipping volume. The article does not support your statement.
The second Micron boise fab hasn't even broken ground yet, they are still working on the first one. So don't expect these things to be completed in parallel.
Some of my family is pretty happy, though, with the job security as they are pretty convinced these projects are all going to take much longer than what's being stated publicly. Micron is saying the first chip from the new fab will be in 2027... though they also predicted it'd be 2026. The date seems pretty slippy.
Anyone who has been around the semiconductor industry since the last century will remember various huge fabs e.g. in Arizona that were partially built but never finished due to oversupply by the time the walls and roof were done.
Capitalism eats itself this way. Second and third order effects will collapse the demand.
You need to keep the market healthy, not some insane Bitcoin style HODL pump - that's how you get wrecked.
I mean I'm not a neoclassicalist but I've read all of them. I'm in consensus with them here. There's a bunch of theories on what a healthy market is but what we're currently seeing matches none of them.
It's short term profitable but long term disastrous, especially in a world where new mathematics and techniques could literally collapse the demand overnight.
Imagine if some paper hits arxiv and the 256 GB requirement for some model now becomes 64. Woops!
Some clever trick about how attention heads and context Windows work could potentially slash a bunch of requirements by giant margins and all they're doing is firing the starting gun at that global race with every obscenely priced unit they sell.
But if prices were reasonable, this wouldn't be an apocalypse. It'd be fine. Consumers wouldn't rush to 64GB, they'd say " Cool I can multitask now at 256" or " great I can do horizontal scalability' or something else.
But no they created the market conditions so now what would happen is the consumer will immediately flip the 192GB they don't need on eBay, hoping to snatch a profit before the prices tank and the second hand market will be flooded the rug will be pulled out from the luxury pricing and everyone will get screwed.
This has happened in electronics markets before. Many times.
When Engels talked about the grave diggers of capitalism they were looking at it through a 19th century labor/manufacturing lens but arguably this same dynamic is at play here.
> Imagine if some paper hits arxiv and the 256 GB requirement for some model now becomes 64. Woops!
If you were a DRAM manufacturer, isn't this exactly the kind of thing that would make you think twice about investing years and $billions in new fab construction?
What "second- and third-order effects" do you suppose will collapse the demand for RAM? The people complaining most loudly about RAM costs are the people who want to run local models; if that becomes popular it will supercharge RAM demand, because locally-hosted models can't parallelize runs from many users the way cloud-hosted ones can. I don't see any slackening in RAM demand at any point in the foreseeable future, even if the big AI companies all go bust.
This is all hypothetical and debating hypotheticals isn't productive so let's roll back to markets.
Let's say ram used to cost $100 and now that same unit costs $1000. You paid say $500x1,000 for that unit during the price increase or some price where you can currently flip for profit.
You have a very expensive data center and you're in debt financed on the premise that you have these special computers.
Now a new technique comes out and it turns out you only need 1 memory unit for something that used to require 8 or 4 or some meaningful multiplier.
This stuff happens all the time. It's why we don't use BMP files on websites or serve giant MOV files on YouTube. It's why postgres queries are faster now than they were 10 and 20 years ago.
You rent out your machines. You need to service your debt.. Demand may 8x overnight to accommodate but you have a monthly bill to pay and that's unlikely. It's likely going to drop.
Think about it. Your customers are paying maybe $10,000 a month and serving their customers. Now they can drop that to $1,250.
On market if you were to sell some of that ram you have 100% profit right now but not for long.
Jevons paradox assumes unlimited capitalization, zero debt servicing, infinite time horizons...
We live in the real world so what do you do?
Historically the answer has been "sell that shit"
There's an aphorism for this "stairs on the way up elevator on the way down"
If we had a healthy market with sane prices where you can't flip the thing you bought for 100% profit the answer would be "create more value."
> locally-hosted models can't parallelize runs from many users the way cloud-hosted ones can
Why not? Unlike many other workloads, LLM inference actually seems pretty suitable for decentralization (effectively stateless means no availability concerns; bandwidth and latency are relatively forgiving too).
I think locally-hosted models at the org level will definitely be somewhat popular, but you seem to be talking about decentralizing for people's personal, non-business use, and I just don't think that's going to happen to any real degree.
People who say they want local runs really mean it: they want local runs on hardware in their room, not on some decentralized system which, if it existed, would almost certainly just be a worse, less-reliable version of cloud hosting. I'm not saying nobody would use it, but it sounds a lot like things like IPFS, which have also completely failed to displace either cloud storage or buying a bunch of disks for your own private use.
>Seriously, if a single politician stepped forward and said "i'll bring down ram prices" they could then shoot a puppy and call me a slur and I'd still go out and doorknock for them.
How many people, outside of tech geeks and megacorps care about RAM prices? And how gullible would you be to BELIEVE the politician they could actually make it happen, and even if they did, that it would extend to the average person, and not JUST megacorps/megadonors?
RAM manufacturers are bidding against NVIDIA and everyone else for the same constrained supply of EUV machines. And it takes years to build more fabs. Micron has multiple fabs coming online in 2027 and 2028.
If we take some time to understand how HBM memory is manufactured (with particular focus on yield risk for final packaging steps), we will hopefully learn that the current capacity crisis is not bullshit.
I guarantee Micron & friends are not intentionally orchestrating their business such that they would suffer a massively reduced chance of yielding on a per-die basis. Unless someone is actually buying HBM devices, they are not going to be making them. These are not a commodity that can be speculatively manufactured in any economically rational way.
What do you think they'll do? Neither repubs nor dems will touch ai companies in a meaningful way. Anything China does wrt memory fabs week be more significant
Are you counting the n-gram/PLE as part of the model weights there? They can go in host memory. Would be good to show your working. Also the released weights are pre-quantised and presumably QATed, so your "Full Precision" and INT8 are simply not a version of the model that actually exists.
Edit: I went and checked for you. The LM backbone is 307.2 GB (286.1 GiB), straight from DeepSeek's upload. The n-gram table is 203.1 GB (189.1 GiB), which goes in host RAM. Note the embeddings are higher precision than the expert tensors, so it's a larger fraction of the bytes than it is of the parameters.
There's no BF16, original full quality weights are quantized already and 510GB.
Then good portion of those weights are n-grams (~200GB) that don't need to be in VRAM.
Then KV cache of that model is super lightweight at ~1GB per 1M tokens. If HBF succeeds, then accelerator with 16GB of VRAM and 1TB HBF/NAND is probably all you need (?).
Projects like DwarfStar https://github.com/antirez/ds4 really lower the hardware bar a lot so Deepseek 4.1 flash and other mixture of expert models can run on consumer hardware. There are also other inference providers who make their money serving openweight models. Services like OpenRouter make it all too easy to utilize these models. Access to these models isn't hard. The hardware moat is becoming pretty easy to bridge.
More concretely DwarfStar M5 128GB Deepseek 4.1 flash 1K tokens @ 29s, 5K tokens + reasoning @ 147s, 10k token prompt @ 463 tokens/s = 22s. Hardware buy-in USD$7K / AUD$8.5K / EUR€6.8K. At typical workloads, ROI is still poor vs. current-era subsidies, but owning hardware is good for privacy/longevity/connectivity independence. Whether you actually consider Apple hardware 'owned' is a valid and thought provoking question.
Still gonna take 2-3 years to get DeepSeek V4.1 Flash quality at decent speeds on reasonably priced hardware.
Hardware update cycles are 2-3 years even on the high end, so it's still a ways away before "good enough" and "local" belong in the same sentence for the average person.
And by then, DeepSeek V6 Flash will be too cheap to meter, 5x faster, and 10x better, so... You'd still need to go out of your way.
Most people are spending most of their time on their phones anyway. ..
"1070s or 1070 TIs because GPUs have been severely overpriced for too long" ... ."
1070ti launch MSRP was $450 ish.
5070 could be had in the last year for 5xx-6xx range easily.
All things considered - (inflation being about 30%~ (guess)) between these two timelines. You are looking at 300% performance difference at a cost dollar for dollar that is cheaper then when they purchased their cards.
Might be a bit of a stretch blaming it on "severely overpriced for too long..."
GLM 5.3 Flash runs fine on two Sparks and Qwen 3.8 Flash Next on one is indeed incredible! I made this 3D game with it in two days using Qwen Code as agent:
Allow a question from someone who’s only got a very vague idea of how this kind of stuff works behind the scenes: say I rent usage of this model through one of the many LLM hosting providers out there, and let‘s assume I use it extensively through something like Pi or OpenCode and vibe code away all the time, keeping the hosted model occupied as much as I can, happily burning my credits.
Does that mean that there is a hardware cluster as described by you above that is crunching away just for me?
So at FP16, I alone keep a 1,664 GiB system occupied all the time?
It depends hugely on what "rent usage of this model through one of the many LLM hosting providers" means. If you're asking them to host the model privately then yes, all of that 1.6T of RAM is likely in use holding weights, activations and KV cache by an inference engine that's only getting/answering requests from you alone. When you aren't actively using the model the hosting process is still active and waiting with all of that memory still wired to it.
As background: For the most part VRAM oversubscription/paging/swapping isn't a thing in the same way that RAM for a VM often is. There are some approaches to it, but (to my knowledge) not at that sort of scale.
There are some systemic reasons for this, but very broadly speaking the GPU vendors are building toward the highest bandwidth and lowest latency possible, and the overhead/complexity of something like protected memory modes serves neither of those priorities.
No, a cluster can server multiple users at the same time, providers cap the tok/s so that one cluster can run inference on multiple inputs at the same time. OpenAI with their new ultrafast mode is probably reserving the whole cluster or prioritizing requests of ultrafast users above others with a higher tok/s hence the high price and high speed. There's many other knobs providers tweak that they don't show the users, for example I doubt many providers are hosting the full FP16 version.
The article isn't just about running locally though. The author is saying it's super cheap to run the model through Opencode Go (and presumably OpenRouter etc.) Personally I'm always most excited by models I can actually run locally, but even these huge open source models open up the competitive landscape for companies to let you call models via an API or just lease compute. And they don't have to charge you to offset research, training, huge staffs of the best minds in the world, crazy PR etc. I think that's a big win for customers and buts competitive pressure on the frontier labs as well.
I did somet math and completely gave up on the idea of trying any worthwhile local model and figured I'd rather pay the 15-30 USD per month via subscription and/or API key combos for years than buying a local setup which might go out of date very fast, if it doesn't goes kaput just out of warranty. I won't be surprised if RAM scarcity is an concerted effort to herd people towards the remote models :)
I reimplemented most of the features of the Deepseek v4.1 flash paper (apart from quantization aware training which doesn't make sense because my implementation uses float32 precision anyways)
I'm currently learning how to distill reasoning traces (check my other github repositories) but I think that a locally selfhostable deepseek is possible with my mixture of experts sharding mechanism. I decided to optimize everything for CPU parallelization, with the idea that the KV cache and meta model have to run from CPU RAM anyways, so the experts can also be loaded/unloaded at runtime if needbe, to save more RAM.
My assumption is that the KV cache optimizations in combination with the CED and compressed attention features are the reason why v4.1 flash has so few hallucination problems and such a strong self-lookup/thinking behavior. But that's more a gut feeling, need to evaluate and test this more thoroughly.
Anyways, would love to see someone train this on their own datasets. Currently my pipeline is kinda optimized for parquet and zim files.
You can run it locally for the price of a decent car, or run it (hopefully) privately on somebody else's hardware at vast.ai or a similar provider for much less. What's not to like?
No, you won't get frontier-level intelligence on a 1070Ti. Yes, it should be illegal to do what Altman did. Since we clearly don't live in the best of all possible worlds, we need to settle, and DS4.1 Flash is a good place to do that.
For tasks that don't require vision I personally like the NVFP4 quant of GLM 5.3 from Local Inference Lab better than DS4.1F, but they are both well beyond awesome.
He doesn't ruin the cost of memory. Advances in memory size and speed are now in full speed mode. Expect drastic increase in the upcoming years. Big factories are in the making and planned. Gigalab in the US and many others in the east.
Since 2010 we have computers with 16gb as being normal. Finally we are moving into a new era where the standard will be 64gb next year and 128 in 2028. Hopefully we reach 1tb in 2030.
I have a pretty large, complex project I've been building with heavy AI use (new language + compiler). I was following a 'strong model as orchestrator launching cheap models as implementers' pattern, but I recently trialled just using Deepseek-V4.1-Flash as the model for both layers because of the cost savings (with mimo v2.6 flash on code review agents for some decorrelation).
I was previously using GLM-5.3 as the orchestrator, after switching to DS anecdotally there was an unnacceptable quality loss, mostly around not taking all the relevant context into account when making decisions, pulling new design out of thin air without discussion too often, and being way too wordy and rambly in documentation despite prompting to avoid it. There's a lot of docs, rulings, core concepts, design philosophy to uphold and DS was just not cutting it.
However, it's perfectly capable of being the sole agent for all of my well specced implementation tasks. I've gone back to GLM as the orchestrator.
On the Claude side of things I was previously following "strong model directs weak" with Fable directing Opus/Sonnet (its choice per-task). Since Opus 5.5 came out I've just been having Opus direct Opus.
The sub-agent separation is still valuable to keep context clean for the orchestrator, but I just have no reason to use Sonnet as the grunt-work implementer because I'm finding it hard to run out of tokens with Opus 5.5 on a $200 subscription plan. It's really really good at subjective quality of work per token used.
I would probably go that route if I could use other harnesses with claude models, but I don't want to be locked in to claude code, and their API pricing (which you need to use it with other harnesses) is so much higher than subscription.
I'm paying for the heavily-discounted subscriptions, not the API rates. There isn't really a cost gap for me. DeepSeek doesn't have a subscription to compare to, but when I compared the GLM 5.3 usage I got from a $100/mo Z.ai subscription compared to Opus 5.5 on a $100/mo Claude subscription, there wasn't a big gap. And GLM 5.3 is very clearly not a frontier model (deepseek v4 seemed a lot
closer, but I didn't use it enough to really say for my workloads).
I don't think those subscriptions nave negative contribution margins, either. I think we're seeing a lot of price discrimination by the big labs, and huge margins on their frontier models. The fact that they have been cutting prices to their second-biggest tier of models (Opus/Sol).
Open models catching up and collapsing these margins would worry me if I were a shareholder in the big labs, but as a user, I really doubt that the western labs have bigger environmental impact just because they have higher API costs, I think they have a ton of efficiencies they aren't sharing with customers yet because demand is so high.
The tightening of subscription value has already begun. dsv4.1f is already worth paying for at market API prices. Maybe it goes to 2x because apparently no one has figured out how to match DeepSeek's insane caching efficiency, but I don't see it getting much worse than that.
Plus you can also get dsv4.1f subsidized. OpenCode Go gives 4x if I understand their pricing correctly. Anecdotally, I feel like I get way more out of my $10/mo OpenCode Go sub for the price than my $20/mo ChatGPT, even using gpt-6.1-sol high which is very cheap, and I have yet to convince myself dsv4.1f is a worse model.
It's really not cheaper than frontier subscriptions. It's getting closer, and it's a great model, but it is not more value per task than the frontier subscriptions. Don't be swayed by the token costs, it's very chatty, like 3x more tokens for the same task as sol. I used dsf 4.1 full time for about a week.
It blows frontier API pricing out of the water, but again, look at cost per task, not token usage. Still easily wins though for my work.
I do think it's the most viable alternative I've seen so far, and that applies pressure to the frontier models. Should subscription prices hike or become unavailable for some reason, I know what I'll be using.
When pricing this, it's important to consider whether or not you want to opt out of data training. You won't get the advertised rate. Also the dsf 4.1 subscription providers are throttled af... and of course they are, because otherwise they'd be haemorrhaging money.
Oh man. v4.1-flash has been an sbolute game changer for us. We run all our Personal Assistants now on flash (thinking high) by default and it works incredibly well. There is really no need for basic agentic tasks that might require Kimi K.3 or GLM-5.3 levels.
Once its gets juicier, we let flash launch specialized subagents with specific models. GLM-5.3 for coding or Kimi K.3 for research and critique.
But as a main driver. I love flash. And it brought our bill down by A LOT :D
What is the cost of access like for DeepSeek-v4.1-flash, compared to GLM-5.3-flash via ZAI's Coding Plan? Because that's what I use; and often hit the "wait". I wouldn't mind trying a new model subscription or even API access which hits around glm-5.3-flash level weight class (which seem to be enough for me; with quite some human suprvision and nudging) but gives muuuuuuch moooore tokens for the same price.
Not shilling for them but Ollama cloud hosts domestically with ZDR afaik. I run 95% of my open weight inference through them. The rest goes through Opencode Go $10 plan (which is enough to run 3 hermes agents on DSF 4.1 and leave plenty of left to experiment with when new models drop).
Just use a provider hosting it in your country especially if your country has major data centers then its the same as using Anthropic or GPT of GCP Model Garden or AWS Bedrock
nobody here is talking about running frontier level intelligence locally so if you’re Chinaphobic and prefer layers of corporations siphoning your data in between you and the party there are plenty of options instead of directly to the party
> Today's models are now good enough for high-quality unattended tasks. Chasing the latest and greatest is silly. It is fun to see the new Fable capabilities, but the tasks we throw at them are usually ridiculous (maybe even insulting) if you believe in LLM sentience. It's like asking a math PhD to organize the files on your desktop.
I'm using DS V4.1 Flash as my main model since their release and it works great for all my coding tasks. My setup is OpenCode Go subscription and obra/superpowers skill.
The only times I try to change models are on general planning tasks (like research this codebase for tech debt mitigation opportunities) or if I need deep research which would benefit from searching the web, in which I still think Gemini is still the best because of the speed and access to google search index. But these are not even 20% of my daily tasks.
I have been using DeepSeek 4.1 flash intensively for over a month. If I run it all day long it costs $1-2. Its fast. Previously I was always quickly running up to my Claude/Codex 5 hour window (on the $20/month plan). The cost savings of DeepSeek is real as shown in this article and I am using subsidized plans.
DeepSeek is horrible at grilling sessions (the /grill* skills to make technical decisions). It doesn't know how to explain things. Maybe the skill could be adjusted. It also doesn't come up with as good solutions as Opus/Sol.
What I use it for is
* the orchestator of my coding workflows
* the tester/verifier of code changes
* the sub agent that explores code or does web searches
* putting together code base research reports
Previously I planned with Opus/Sol/Astra and then I used DeepSeek for coding, and then reviewed with Opus/Sol/Astra. With the cost improvements to Opus/Sol I am trying to use them for coding instead now so there will be less back and forth review needed.
They are all working together in Pi using the extension @tintinweb/pi-subagents where my workflow skill is calling different subagents that use different models.
Luna is cost competitive, but doesn't score as well on intelligence. I do need the intelligence for most of what I use it for, so I am not motivated to use Luna. Haiku also doesn't seem like a competitive price/performance mix.
It is super bad on a bit more complex workflows and starts repeating same errors with the same tool until the cycle breaker hits.
6 is worse than 5.6 here.
But it is amazing on generating a report on content generated by better agentic models such as DeepSeek or GLM, which both do a mediocre/bad job on reports.
DeepSeek's own paper advises against using Max, showing that it normally doesn't perform that much better. I am not using it on Max, so that's not a useful benchmark for me. I have seen other benchmarks where Flash does significantly (30%) better than Luna.
Because DeepSeek is not "a month or two" behind as claimed in the article.
These open models still did not beat February's Mythos / Fable 5.
DeepSeek 4.1 Flash is behind GPT 5.6 Sol, and that one is left in the dust by the excellent Opus 5.5.
Rumors say Anthropic is holding in reserve the big improvement, Fable 5.5, for the IPO.
It's plausible that open models are 6 - 12 months behind, and there is no "good enough". As long as progress doesn't slow down, leading labs have nothing to fear.
Many people would, and you'll find that they're building crappy webapps where you dont need SoTA. Like seriously who needs these frontier models?
Unless you're doing some extermely difficult post-grad lvl research, you do not need a 100x PhD research assistant, especially not for whatever silly SaaS product most people are building.
There's people at my job that get so much more done than everyone else using Fable. and all they use is the fastest cheapest models. I'd say the people who are using sota models for everything are doing it just because they prefer to be lazy.
You simply do not need these frontier models, they outgrew most people's needs 6 months ago, but for some reason people still want to run a 700k rack of gpus full throttle to center a div for them.
I am coding CRUD apps with a mix of astra, sol 6.1, fable and opus 5.5. A more capable model would still benefit me imo. Being able to follow high level guidance better, and being able to harness other models for each task would be a big improvement.
There absolutely is "good enough" and I agree with this author: DeepSeek 4.1 Flash is plenty good enough for all the things I would trust an AI to do at my job.
I am 4.1 maxxing on commandcode GOAT Plan + api rates with oh my pi for the last 4 weeks, it's absolutely amazing and crazy fast, it's alright if it makes a mistake, I have enough time to iterate again, I have also added an advisor layer of mimo 2.6 pro which does make it a notch smarter. Getting haiku 5.5/sonnet5.5 to work on plans and letting 4.1 flash work through it is helping a ton too.
I am a big ChatGPT fan, all our team has ChatGPT Subs, but the TPS across all models including luna is just so damn slow.
Commandcode giving 60$ worth of Deepseek for 10$ is just genuinely goat.
And it never says no for cyber tasks so that's a big win
I had the same experience using ds 4.1 last couple of weeks. It’s insanely good for the price. I’m doing mostly web dev it excels at everything I throw at it. The pricing is ridiculous. I canceled my gpt subscription and haven’t looked back hope the pricing stays like that. I almost never need a better model. I still keep my Claude 20$ sub for now but I feel like one more iteration and I won’t need even that anymore I hardly use it
If DS4.1 impresses you I would be really interested to see your comparison to GLM 5.3. I switched from the one to the other and even if GLM 5.3 is a bit slower I don't think I'll be going back.
DS4 (not 4.1) crossed my dont-care threshold and I genuinely stopped paying attention to new models. I'd love to try GLM 5.3 but I just don't see any point in spending the effort any more. I can get passable intelligence for a bargain price either direct from China or from a ZDR EU provider for a small markup. Paying 10x more will not make me 10x happier, it's unlikely to make me even 1.1x happier now I've got some intuition for the natural limits of these models.
I don't even bother checking how much I spent on API any more, its well under $30 over the past 2 months despite daily constant use. Who even needs a subscription at these numbers?
Been using Flash 4.1 via the ante harness to blast through a GBA recomp. The ante team has pushed hard to make Flash 4.1 perform well under it. So far, I've maybe spent $10 over the last 3 days. Its a real workhorse and works much better in this harness
I did a test on this a couple weeks ago. What I found was that the chinese models were far better than API rates, but about comparable on price vs the subsidized subcription model (chatgpt). Also it was my experience that codex completed tasks quicker.
That said, it's my best understanding that these american companies aren't profitable and will eventually raise rates (the old uber trick) so I'm keeping myself ready to switch when that day comes.
The question seems rhetorical but I think there are two reasons in some combination. First is it there is some awareness lag here. That lag can be on the producer and consumer side. Software enterprises are pretty slow to adopt new things and slow to try new things so they might only be aware of openai and Claude as options. Plus there are some scariness because deep seek is a Chinese model and therefore export restricted - never mind that there are American in European providers.
The other reason is more interesting. Maybe the frontier providers think that price performance is irrelevant in light of very powerful frontier models that can start the RSI loop and or a huge displacement of work and a winner take all economic situation. After all if frontier providers earn everyone's money then you won't have any money to spend on any model 100x cheaper or not.
I think it also has a bit to do with the AI sector of tech still moving at lightning speed.
Theres already models that outdo DS 4.1 flash in cost/performance. Luna 6 on max effort for example. Luna also doesn't care what time of the day it is for cost calculation.
And I'm sure by the time people ask why Luna 6 is being slept on there will be another cost/performance king
Luna is very slow and bad at agentic tasks. DS runs circles around it and there are US providers providing cheaper rates no matter the time of the day.
While using DeepSeek v4.1 Flash I was architecting a system and I made a mistake of drawing the RPC boundaries at a wrong place that did cost me in so many ways.
I realized that mistake and guided DeepSeek where it should be.
Next I fired Fabble 5.5 set to high to check if the hype is real about Fabble. It exhausted 89% of quota and came up with NOTHING that DeepSeek hadn't flagged itself already in its notes.
Do you mean Fable 5.1? Or Opus 5.5? I'm not sure what you're working on but for me DS 4.1 flash isn't nearly at their level. For the price it's obvious very impressive, though Luna 6.0 is excellent too.
The problem with benchmarks and proprietary models is that one day a model is best at doing X, another day that's not so sure. And anyway, we are not throwing the same X.
I've found supposedly smaller and, less performant models do better on certain tasks. I end up using several models, sticking to what my unconscious statistical observations tell me to use for the kind of task at hand.
> With my OpenCode Go sub of $10/month, DeepSeek is basically unlimited
My OpenCode Go monthly window was scheduled to reset this morning. It was sitting at 22% used despite me using DeepSeek V4.1 Flash heavily as my implementation agent the past couple weeks (I use gpt-6.1-sol high for planning/orchestration).
I had 1.5 hours left so I fired up first 10, then 20, and finally 50 concurrent subagents all working on reverse engineering C code from an old PC game. They found over 100 new functions.
This is the first workload I've found that could make a dent in my sub. It got my 5 hour window to 85% used, but sadly my monthly was still only at about 35% when it reset. So that cost maybe $2.
How are you guys doing orchestration? I have been fumbling around in my free time trying to build something for myself but is there a repo or something that just works?
I might not be the best person to ask. I use Pi harness in tmux and just ask my current agent to spawn interactive pi instances in new tmux windows, create a sentinel file for each of them, and monitor the sentinel files for signals every 2 seconds.
Currently have auto compaction turned off. When the orchestrator's context is getting close to full, I have it write a handoff markdown file and point a fresh agent at it.
I do feel like I'm getting close to the point where I might be ready for something more sophisticated, especially wrt to subagents communicating with the orchestrator.
I think the interesting provider to cross-check this assumption with here is Meta, who is clearly freaking out, and is currently providing Muse 1.3 even cheaper so long as you are willing to share data with them
I think because we're all just using it thinking we have found the "model for me" and never mentioning it to anyone because what would we say? It's good. It's a bit like the Logitech MX Master, as more and more people assumed they had found the ideal mouse for their purposes, it quietly became the professional standard through sheer adoption.
Technically yes, but has been reported to be quite benchmaxxed. In practice Deepseek Flash 4.1 and GLM 5.3 might therefore still outperform Mimo 2.6 pro.
I’ve been using it a lot and it’s performing really well. Not GLM 5.3 levels but it beats Deepseek for my use. I’ve used it on long running coding tasks, though mostly prototyping, but it’s done a great job at very low cost.
Enterprise is not freaking out because DeepSeek 4.1 Flash does not actually occupy a spot on the Pareto frontier for non-coding enterprise workflows. We see this at my employer, focused on non-technical knowledge work. Luna 6 and now Haiku 5.5 are both very competitive if not better on all axes that we care about
Deepseek V4.1 Flash is a hidden gem, really. Not to mention, you can easily get it through many providers that offer zero data retention and consistent speeds above 200 tokens per second!
It's a solid little model, and I appreciate DeepSeek's commitment to the bit in releasing a brand new pretrain, double the size, numerous architectural innovations as a ".1" release over the excellent DeepSeek V4 Flash.
Yeah. $100 for Claude just about gives me all the usage I want, as a more or less full-time hobbyist having it work in the background most of the day. I was trying to economize by having a local LLM, then Deepseek, then Cursor/Grok, and then I got a taste of Opus 5.5 and I simply cannot go back to having to carefully spec things out and double-check work. I just let it decide, Opus or Sonnet for the next task, and I get almost perfect results. Probably similar with OpenAI's models.
The token-equivalent monthly spend is > $5K+. If Deepseek's token cost is 20x cheaper, that's $250/mo, and I'd be spending a lot more of my brainpower babysitting it and getting worse results.
For business/team accounts that pay per-token, maybe I can see the "freaking out" being warranted on the part of the fronter labs. But as long as they're willing to subsidize their end-user subscriptions, I'm not going to move off of them until the alternatives are truly at their level.
Agree that people leaving the big two companies is going to be hard to keep a pulse on prior to IPO.
Anthropic and OpenAi are in the news, so they get the press and people go and try out their product. Large enterprise businesses are going to make larger, longer-term contracts with them and are only going to pivot if they think switching costs are easy or if they think the provider won't deliver.
The other inference producers are less well known or you need to get your cloud sales rep to tell you how to switch to them as a provider rather than Anthropic or OpenAI.
I use OpenRouter, I know switching is easy, but larger businesses tend to work in yearly cycles. DeepSeek v4 Flash came out in late April.
I agree OpenAI and Anthropic are going to struggle when the median price of running a smart-enough model keeps falling.
Edit: I also think demand for hardware will be rapidly absorbed by other companies if Anthropic or OpenAI stumble. We've finally turned hardware directly into runnable intelligence and people are not going to go back to the old ways.
People tend to conflate the question "is AI a useful technology?" with "are the AI companies going to do well?" but they're surprisingly separated in practice, with either one able to be true while the other is false. There is a lot of money tied up in a lot of hardware with a lot of loans made against that hardware as collateral all based on the assumption that AIs are going to need more and more and more and more hardware and whoever has the hardware wins. If a much better model comes out that requires vastly less hardware, or even more accurately, merely charges vastly less than the current AI companies, then to a first approximation (barring Jevon's paradox, and bearing in mind there's no timeline guarantee on that) all that hardware becomes much less valuable for being grotesquely oversupplied relative to what is necessary, and even though that would generally make AI objectively more useful than it was before, it would cause mass financial chaos in the markets.
The markets need a very particular rate of progress. It isn't entirely clear to me that it's even a possible rate of progress, it may be overconstrained, but they certainly don't have plans for the AI models to get commoditized on the timeframes of these vast, vast array of loans being made against hardware as collateral. Spend a metric shit ton of money to kill all your competition then charge monopoly rent on the one thing absolutely everyone needs doesn't work if you can't economically "kill all your competition" because the economics favor them in the spending spree.
And then, based on the fact that this is not even remotely complicated logic, there are plenty of people who are fully aware that they have a lot of money tied up in not running around telling everyone how wonderful the cheap models have become.
It's also unclear whether those who approved those loans understand GPUs depreciation. In any case, progress in software but also hardware could bring chaos and ruin their house of cards.
> It's also unclear whether those who approved those loans understand GPUs depreciation.
This also assumes heavy utilization, though. If there's heavy utilization, it might mean they're doing well. If they're all spinning, it's time to raise prices.
they should be freaking out because every time the chinese labs or non "frontier" labs release a model that is only a few months behind and much cheaper than the openai/anthropic models it shows that they don't deserve their valuations
Perhaps, OR it might be that most people in the markets (think that they) are not all that exposed to the valuation AI labs and so their eventual collapse doesn't matter.
Or perhaps they consider the upside from cheap Chinese models to hedge the effect that OpenAI/Anthropic collapsing would have on their portfolios. This would make sense for (hedge funds holding) most companies: they don't really care about who supplies the AI, as long as they get it at roughly the same price as their competitors.
ironically, at my large enterprise, they aren't yet distinguishing between "chinese models" and "chinese models hosted at microsoft foundry". so far it's just _banned_. i'm not at all pretending it's like that at other orgs.
I have used over 40B tokens and spent over $800 on DeepSeek API over the past 30 days, mostly on V4.1 Flash.
It's good, and you can do most work with this. For complex software implementation you need to split your runs into various phases, build in verification, and use subagents so that work gets another audit and repair pass from the lead agent. You can do pretty much everything then. Frontier models can do without compelx workflows, that's the difference.
I built mjolnir in large part so I could have Opus manage DeepSeek Flash subagents. It's phenomenal and extremely light on the Claude tokens. https://github.com/BrokkAi/mjolnir/
And yes, Opus is enough smarter than DSF that it's worth the extra steps. This ranking is from live tickets, no contamination: https://slopcop.com/power-ranking
Opencode Go gives only 6300 requests for glm and 23 000 for deepseek. And, if I wanted to, I would be able to do all my work on $10 plan with deepseek. It’s very cheap.
That is opposite to my experience so far, can you describe your coding tasks? Mine are systems level code, utilities, operating system code, networking and real time control stuff.
I also prefer GLM-5.3-flash to DS-4.1-flash, but it's close. Since Z.ai has been offering essentially free GLM-5.3-flash tokens on their coding plan between 8am-6pm pdt I've been using it a lot... though that ends on Oct 10 IIRC.
DeepSeek 4.1 Flash 0910 is perfect for M5 Ultra 256GiB. Running it fully resident in RAM, prefill at ~2500 tok/s and decode at ~40 tok/s. Probably tons of room to improve from there.
Don't you run into it sometimes outputting a few Chinese characters, or Cyrillic, for no apparent reason? I fear it writing some nonsense in the code or the terminal. DeepSeek V4.1 Flash doesn't seem to do that.
What blows me away about this model is it's speed.
It's way faster than Opus or any of the GPT models.
I have a coding harness which is opencode plus a few skills relevant to my workflow. Deepseek 4.1 Flash does very well in this environment. I haven't noticed much difference quality wise compared to Opus 5, which I use in my day job as my employer pays for it (although I'm considering using DeepSeek here too given how cheap it is).
It's still lacking llama.cpp support, and the work on that isn't moving very fast either. Looks more like a general community issue, where this model isn't drawing much interest.
People have been raving since forever about Deepseek, but if one looks at the CoT, it's evident that it's way way stupider than frontier models (there's a reason why it's cheap). It's laughable to compare Deepseek 4.1 with Opus 5.5.
I've benchmarked, rigorously, deepseek-v4-flash for programming and personal use, and it is definitely less smart than Qwen3.8-flash-next (which in turn, is not terribly smart).
Local models are also really slow, unless one spends insane amounts of money.
Having said that, Qwen3.8-flash-next is an impressive evolution; it reaches the small versions of the frontier models (like Sonnet) - but again, it's massively slower and not 100% reliable (including: stability).
The argument that most people are making isn't that dsv4.1f is better than frontier, but that it's good enough for most tasks, faster, and way cheaper.
> if one looks at the CoT, it's evident that it's way way stupider than frontier models
Honestly for me the intelligence gap between DS 4.1 Flash and Muse Spark 1.3 makes Muse more worth it for me, especially on a $10 OpenCode Go sub, with the caveat that everything I use it on is open source which makes the fact that I'm sharing it with Meta a little moot because it's already published permissively on GitHub anyways.
I'm no expert, but it think that it's the pricing on GPT-6 Luna. I'm also guessing that it's been underpriced just for this reason. I also don't think it's all that great, but it's definitely very cheap.
If it's underpriced, it's a loss leader to sell the other models, so it actually can't be too good.
I really put these things through their paces because I use them to review and work with new abstract game rules and models, so they're always flying blind. Luna misses the obvious (and more importantly, the clearly explained) consistently. My second prompt is listing all of the points in its first response, and saying "No, it doesn't work like that." The third prompt is picking out the two or three suggestions it made after correcting itself on all of the original points and saying "That's how it already works." The fourth prompt is "Now that we're done going over the rules, can we start?"
In my experience it just takes so much longer to arrive at "done" state for me. It thinks for soooooo long. I guess if you're running 12 sessions at once you don't really notice.
if you have a legitimate coding application, it isn't very good. if you have some kind of inauthentic activity, which could be what it is trained for for all sorts of reasons...
What is your evidence? Deepseek v4.1 Flash is by far the most popular coding model on openrouter, having processed 38.7T tokens in just the last 7 days, over 3x the usage of the 2nd rank model.
So I ask again, what are you basing your assertion on?
my own usage of deekseep v4.1 flash, and that among the dozens of great programmers i know, not a single person is using it
BUT. they are employed to do / deciding-to-do authentic (if often meaningless) stuff.
here's a short list of inauthentic activity that claude and openai refuse to do:
- chat services that, when you ask them, say they are not chatbots when they are
- code to work around software licenses or DRM
- code to scrape or download copyrighted material
- directly cheating on homework
- adopting a persona in social media that spreads misinformation or propaganda
this is but a short list. but ask me, "are there enough inauthentic activity demands such that someone who CANNOT USE claude or gpt as the LLM would use dsv4.1 on openrouter instead?" yes. i mean there are whole countries right now where the culture can be summarized as, "bottom to top, inauthentic activity." i am surprised it is not more usage!
So your argument is that the 38T of tokens used in the last 7 days is by moron programmers or people doing "inauthentic" tasks? You think the person who made this post is also an idiot?
Do you realize how incredibly delusional/self-centered you sound?
AA shows Luna at 1/4 the price, 1 point behind on intelligence matrix with a 38.
Haiku 5.5 is 23% cheaper with a 4 point intelligence lead.
I'm on subscription usage so I can't compare Flash 4.1 to them directly but the OP has his head up his ass if he thinks Opus 5.5 is the best point of comparison. Why is anyone using Opus if the new Haiku is indistinguishable /s
Just absolutely terrible post, admits to using Opus for review but claims its intelligence isn't needed, why aren't you using Haiku or Sonnet then?
You might also choose to pay money for a service that provides real value instead of actively choosing to support the Chinese deliberate effort to undermine this country.
The reason people aren’t freaking out is because most people are using heavily subsidized subscriptions.
I tried the cheapest provider on openrouter and burned through $50 in a few days. Quality was ok, seems slightly above Luna quality perhaps? But that $50 is 1/4 of my codex subscription where I could have burned that many tokens or more using Astra within my weekly reset.
This won’t last forever but as long as the frontier labs are subsidizing this heavily the open models won’t matter.
Yes. You have to find the provider with pricing that suits your usage.
I am having 98% my input in cache, so using Coralbricks makes sense due to them giving cache reads for free — you only pay for writes. I spend maybe 5-10 dollars a day and my agents basically work day and night implementing things for me.
If your tasks are write-heavy, find a provider with cheaper output.
If you build a customer-facing app, pay a bit extra for 400+ tok/s e.g. on Lithos.
But aren't you developing bad habits and learning patterns that won't work long term? Or do you think things will get cheap enough that you will be able to keep going with your current patterns post-subsidies?
I expect by that point we'll have local models that can do a decent job, I would guess give it a decade and we'll be running custom accelerators that are smarter than current frontier models.
In the same way that only supercomputers used to have multiple processors and caches but it's now standard.
> But aren't you developing bad habits and learning patterns that won't work long term?
2 reasons - there's an advantage now, use it. 2nd the frontier providers, this is the "early cheap days" like when uber was initially cheap to compete vs standard cabs. they want you to become hooked and boy are we hooked.
No one involved in this is thinking about the long term
How's the caching? I have 99.5% cache hit rate with deepseek when using their own API, it's dirt cheap.
I am curious how you managed to spend that much on Deepseek via OpenRouter. I loaded $100 back in July while using v4-flash or whatever the cheap good model was at the time, and have upgraded as the new ones came out from Deepseek. I still have $16 and some of that spend also goes towards the AI usage from my customers (the context they need to load in is quite large too).
And I am using the Claude Code harness with DS as the endpoint. And I use it ~5-8hrs a day to do my coding.
Likely the user doesn't know what they're doing or has extermely bad workflows. They're prob not managing their cache, and dont use compaction.. Letting context get to 500k and invalidating their cache every 10 tool calls because they have no providor fallback settings.
I was running deepseek v4.1 pretty much non stop during work hours, with heavy tool/mcp usage and finding it very difficult to spend more than $75 in a month.
Also the cheapest providers on Openroutrr can often have terrible cache hit %, short TTLs resulting in their effective price being much more expensive than people realize. 75% cache pretty much destroys any savings from a super cheap token perspective.
OpenRouter is complete garbage.
Buy directly from DeepSeek's API.
You can literally get overcharged 100x on DeepSeek on OpenRouter (or more).
I'm all in for saving money and _can_ move to using DS directly from them, but maybe I am missing something here:
OpenRouter Pricing:
$0.02/M input tokens $0.60/M output tokens
DeepSeek Pricing (cache miss, off-peak):
$0.15/M Input $0.60/m output
When 98.5% of my requests are cache hits (according to Pi for the last week), the cache miss price isn’t that important to me, and $0.003-0.006 per 1M input tokens is shockingly cheap.
It’s also the major difference between using DeepSeek directly vs other providers also serving it, though I have not looked lately: it’s possible other providers have matched its cache hit pricing better?
Interesting, if the cache hit is that good, I think HN convinced me to toss $20 at DS official, and see how long that lasts.
I've heard that certain inference providers may have different quality of caching implementations, so even if the listed numbers are as you say, the practical cache hit % you get might be significantly different/incur significantly different costs.
[delayed]
Which provider was that?
Also DeepSeek usage is subsidized as well, it’s a power hungry model.
Interesting. Are all of the providers on OpenRouter simply losing money? How does that even work out?
No, it’s outrageously profitable above x% without stealing any prompts. Provider economics still pretty good. Acquiring hardware is the current limiter.
You're the RLHF.
Not if you're not giving feedback.
With ZDR-only enabled? That seems illegal.
Don't ask and dance as long as the music keeps playing.
Are you sure about that? My impression was most providers on openrouter were purely selling tokens for profit...
Have y'all tried an Ollama Cloud subscription? Their off-hours pricing for V4.1 Flash is extremely competitive.
Call me crazy but:
VRAM & Memory Requirements by Precision
• FP16 (Full Precision): Requires ~1,664 GB of VRAM (e.g., an 8x B300 288GB cluster).
• INT8 Quantization: Requires ~832 GB of VRAM (e.g., 8x H200 141GB).
• INT4 Quantization: Requires ~416 GB of VRAM (e.g., 8x A100 80GB)
VRAM aint cheap, Sam Altman ruined the cost of memory, Nvidia doesnt make enough consumer GPUs letting the market go insane over them, I still have friends on 1070s or 1070 TIs because GPUs have been severely overpriced for too long. I remember when a gaming PC was only $1000.
Even so why would anyone not sleep on a model they cannot run?
Seriously, if a single politician stepped forward and said "i'll bring down ram prices" they could then shoot a puppy and call me a slur and I'd still go out and doorknock for them.
Memory companies have price fixed multiple times. They've paid hundreds of millions in fines. wikipedia even has a page on it. https://en.wikipedia.org/wiki/DRAM_industry_price_fixing.
Look at the financials of these companies, they're all making obscene margins and do they plan to increase production? No. Micron is doing a stock buy back to pump the price of their share.
The Micron CEO just recently said this is the exact plan https://www.theregister.com/systems/2026/10/01/ram-supply-se...
There's sanctions, tarrifs, and a DOJ who doesn't give a shit. Until we can fix that the insanity will continue. Phones will be unaffordable. Laptops will be obscene. Gaming consoles will be thousands of dollars. Desktops will be dead.
If you're waiting for some David Ricardo equation to happen, tough cookies, it's not coming.
The market is legally locked down and we're in hostage pricing mode.
And what's the story? You can't afford electronics because we're using it to build robots to take your job? I mean ...
Nobody is coming to save us. That's our job.
> do they plan to increase production? No.
Micron has 3 brand new fabs currently under construction, 2 Boise, 1 in New York as the first of 4 planned for a campus.
Plus expanding other existing facilities.
These things take ~3-5 years from breaking ground to full production. You'd have had to anticipate the current demand years before it happened in order to be bringing production on-line before 2030 or so.
Samsung and HK Hynix also have fabs under construction and planned.
CXMT started 11 years ago and only now is reaching any real volume. If they decided a year ago to react to the current demand cycle they'd be 6-7 years out.
Not much you can really do to wish for more fabrication to exist on any timeline not measured in fractional decades.
Could they do more and react quicker? Probably, but everything I've read on the subject seems to point to 3 years is absolute bare minimum if you happen to have a shovel ready project with the land bought, local permitting completed, infrastructure extended to the site, and a skilled workforce already in place. They could suspend buy-backs/dividends today and dump it all into building production and there would be no material impact until around 2030.
> The Micron CEO just recently said this is the exact plan
CEO simply stated the demand pressure will not go away through 2027, and supply will not increase until around 2028 when currently under construction fabs start shipping volume. The article does not support your statement.
Stanford tracks RAM prices in this nice little site: https://dam.stanford.edu/memory-prices.html
Costs did go nuts, but there are signs of easing in the market of late. CXMT is starting to have an impact and priced will probably fall in 2027.
https://pcpartpicker.com/trends/price/memory/ is better. pcpartpicker has the data.
The second Micron boise fab hasn't even broken ground yet, they are still working on the first one. So don't expect these things to be completed in parallel.
Some of my family is pretty happy, though, with the job security as they are pretty convinced these projects are all going to take much longer than what's being stated publicly. Micron is saying the first chip from the new fab will be in 2027... though they also predicted it'd be 2026. The date seems pretty slippy.
Anyone who has been around the semiconductor industry since the last century will remember various huge fabs e.g. in Arizona that were partially built but never finished due to oversupply by the time the walls and roof were done.
Yeah, but why would they make consumer memory when HBM for GPUs is much more profitable?
Capitalism eats itself this way. Second and third order effects will collapse the demand.
You need to keep the market healthy, not some insane Bitcoin style HODL pump - that's how you get wrecked.
I mean I'm not a neoclassicalist but I've read all of them. I'm in consensus with them here. There's a bunch of theories on what a healthy market is but what we're currently seeing matches none of them.
It's short term profitable but long term disastrous, especially in a world where new mathematics and techniques could literally collapse the demand overnight.
Imagine if some paper hits arxiv and the 256 GB requirement for some model now becomes 64. Woops!
Some clever trick about how attention heads and context Windows work could potentially slash a bunch of requirements by giant margins and all they're doing is firing the starting gun at that global race with every obscenely priced unit they sell.
But if prices were reasonable, this wouldn't be an apocalypse. It'd be fine. Consumers wouldn't rush to 64GB, they'd say " Cool I can multitask now at 256" or " great I can do horizontal scalability' or something else.
But no they created the market conditions so now what would happen is the consumer will immediately flip the 192GB they don't need on eBay, hoping to snatch a profit before the prices tank and the second hand market will be flooded the rug will be pulled out from the luxury pricing and everyone will get screwed.
This has happened in electronics markets before. Many times.
When Engels talked about the grave diggers of capitalism they were looking at it through a 19th century labor/manufacturing lens but arguably this same dynamic is at play here.
> Imagine if some paper hits arxiv and the 256 GB requirement for some model now becomes 64. Woops!
If you were a DRAM manufacturer, isn't this exactly the kind of thing that would make you think twice about investing years and $billions in new fab construction?
What "second- and third-order effects" do you suppose will collapse the demand for RAM? The people complaining most loudly about RAM costs are the people who want to run local models; if that becomes popular it will supercharge RAM demand, because locally-hosted models can't parallelize runs from many users the way cloud-hosted ones can. I don't see any slackening in RAM demand at any point in the foreseeable future, even if the big AI companies all go bust.
This is all hypothetical and debating hypotheticals isn't productive so let's roll back to markets.
Let's say ram used to cost $100 and now that same unit costs $1000. You paid say $500x1,000 for that unit during the price increase or some price where you can currently flip for profit.
You have a very expensive data center and you're in debt financed on the premise that you have these special computers.
Now a new technique comes out and it turns out you only need 1 memory unit for something that used to require 8 or 4 or some meaningful multiplier.
This stuff happens all the time. It's why we don't use BMP files on websites or serve giant MOV files on YouTube. It's why postgres queries are faster now than they were 10 and 20 years ago.
You rent out your machines. You need to service your debt.. Demand may 8x overnight to accommodate but you have a monthly bill to pay and that's unlikely. It's likely going to drop.
Think about it. Your customers are paying maybe $10,000 a month and serving their customers. Now they can drop that to $1,250.
On market if you were to sell some of that ram you have 100% profit right now but not for long.
Jevons paradox assumes unlimited capitalization, zero debt servicing, infinite time horizons...
We live in the real world so what do you do?
Historically the answer has been "sell that shit"
There's an aphorism for this "stairs on the way up elevator on the way down"
If we had a healthy market with sane prices where you can't flip the thing you bought for 100% profit the answer would be "create more value."
>Your customers are paying maybe $10,000 a month and serving their customers. Now they can drop that to $1,250.
Or they could stay at $10,000 per month since they are willing to pay that much already.m, so they just use AI more and in more places.
> locally-hosted models can't parallelize runs from many users the way cloud-hosted ones can
Why not? Unlike many other workloads, LLM inference actually seems pretty suitable for decentralization (effectively stateless means no availability concerns; bandwidth and latency are relatively forgiving too).
I think locally-hosted models at the org level will definitely be somewhat popular, but you seem to be talking about decentralizing for people's personal, non-business use, and I just don't think that's going to happen to any real degree.
People who say they want local runs really mean it: they want local runs on hardware in their room, not on some decentralized system which, if it existed, would almost certainly just be a worse, less-reliable version of cloud hosting. I'm not saying nobody would use it, but it sounds a lot like things like IPFS, which have also completely failed to displace either cloud storage or buying a bunch of disks for your own private use.
>Seriously, if a single politician stepped forward and said "i'll bring down ram prices" they could then shoot a puppy and call me a slur and I'd still go out and doorknock for them.
How many people, outside of tech geeks and megacorps care about RAM prices? And how gullible would you be to BELIEVE the politician they could actually make it happen, and even if they did, that it would extend to the average person, and not JUST megacorps/megadonors?
RAM manufacturers are bidding against NVIDIA and everyone else for the same constrained supply of EUV machines. And it takes years to build more fabs. Micron has multiple fabs coming online in 2027 and 2028.
> "i'll bring down ram prices"
wonder what voting would be like?
gamer vote ++
datacenter hater vote --
datacenter lobby ++
micron lobby --
Yep, that's all the voting blocs.
Wouldn’t cheaper memory make it easier to bring compute out of data centers and onto consumer hardware?
Datacenter haters will read this as "that's still evil AI", and everyone else hopefully can count and understands it'll be worse for environment.
How many people would you allow them to kill?
If we take some time to understand how HBM memory is manufactured (with particular focus on yield risk for final packaging steps), we will hopefully learn that the current capacity crisis is not bullshit.
I guarantee Micron & friends are not intentionally orchestrating their business such that they would suffer a massively reduced chance of yielding on a per-die basis. Unless someone is actually buying HBM devices, they are not going to be making them. These are not a commodity that can be speculatively manufactured in any economically rational way.
> they could then shoot a puppy and call me a slur
I know it's just a figure of speech, but damn. I laughed out aloud in public just reading this.
Neither political party cares at all about memory pieces get real lol
I don't really understand why. Memory is a critical component of every computational device.
wait until holiday shopping...it affects the price of almost everything with a battery or power cord.
What do you think they'll do? Neither repubs nor dems will touch ai companies in a meaningful way. Anything China does wrt memory fabs week be more significant
Are you counting the n-gram/PLE as part of the model weights there? They can go in host memory. Would be good to show your working. Also the released weights are pre-quantised and presumably QATed, so your "Full Precision" and INT8 are simply not a version of the model that actually exists.
Edit: I went and checked for you. The LM backbone is 307.2 GB (286.1 GiB), straight from DeepSeek's upload. The n-gram table is 203.1 GB (189.1 GiB), which goes in host RAM. Note the embeddings are higher precision than the expert tensors, so it's a larger fraction of the bytes than it is of the parameters.
So,
> Call me crazy but:
You're crazy. :-)
There's no BF16, original full quality weights are quantized already and 510GB.
Then good portion of those weights are n-grams (~200GB) that don't need to be in VRAM.
Then KV cache of that model is super lightweight at ~1GB per 1M tokens. If HBF succeeds, then accelerator with 16GB of VRAM and 1TB HBF/NAND is probably all you need (?).
Projects like DwarfStar https://github.com/antirez/ds4 really lower the hardware bar a lot so Deepseek 4.1 flash and other mixture of expert models can run on consumer hardware. There are also other inference providers who make their money serving openweight models. Services like OpenRouter make it all too easy to utilize these models. Access to these models isn't hard. The hardware moat is becoming pretty easy to bridge.
More concretely DwarfStar M5 128GB Deepseek 4.1 flash 1K tokens @ 29s, 5K tokens + reasoning @ 147s, 10k token prompt @ 463 tokens/s = 22s. Hardware buy-in USD$7K / AUD$8.5K / EUR€6.8K. At typical workloads, ROI is still poor vs. current-era subsidies, but owning hardware is good for privacy/longevity/connectivity independence. Whether you actually consider Apple hardware 'owned' is a valid and thought provoking question.
Still gonna take 2-3 years to get DeepSeek V4.1 Flash quality at decent speeds on reasonably priced hardware.
Hardware update cycles are 2-3 years even on the high end, so it's still a ways away before "good enough" and "local" belong in the same sentence for the average person.
And by then, DeepSeek V6 Flash will be too cheap to meter, 5x faster, and 10x better, so... You'd still need to go out of your way.
Most people are spending most of their time on their phones anyway. ..
In the words of a Scottish comedian, "The average person is Chinese." https://youtu.be/LsEhNMy8svo?t=98
"1070s or 1070 TIs because GPUs have been severely overpriced for too long" ... ."
1070ti launch MSRP was $450 ish. 5070 could be had in the last year for 5xx-6xx range easily.
All things considered - (inflation being about 30%~ (guess)) between these two timelines. You are looking at 300% performance difference at a cost dollar for dollar that is cheaper then when they purchased their cards.
Might be a bit of a stretch blaming it on "severely overpriced for too long..."
Not quite: not all of this needs to be in VRAM
It has a set of n-gram tables which you can stream from system RAM or even NVMe
That said it’s still quite big! I can’t fit it on my DGX Spark, though I believe you can if you have two?
It needs 3-4 Sparks to run well (at an acceptable quantization and sufficient KV cache):
https://github.com/christopherowen/spark-ds41f
Ah that’s a shame. GLM 5.3 Flash is honestly as good IMO and can run on two pretty successfully from what I understand.
I’m quite spoiled with how good Qwen 3.8 Flash Next is on a single spark though: shocking how good local models are getting on attainable-ish hardware
DeepSeek V4 Flash runs well on two Sparks, I documented that here:
https://blog.jonathanpage.com/
GLM 5.3 Flash runs fine on two Sparks and Qwen 3.8 Flash Next on one is indeed incredible! I made this 3D game with it in two days using Qwen Code as agent:
https://games.jonathanpage.com/
I have access to two and will explore this the coming weeks.
Also give GLM 5.3 Flash a try: it’s shockingly good too in my testing, and I believe eugr has a TP=2 recipe to use for sparkrun
Thanks for the data!
Allow a question from someone who’s only got a very vague idea of how this kind of stuff works behind the scenes: say I rent usage of this model through one of the many LLM hosting providers out there, and let‘s assume I use it extensively through something like Pi or OpenCode and vibe code away all the time, keeping the hosted model occupied as much as I can, happily burning my credits.
Does that mean that there is a hardware cluster as described by you above that is crunching away just for me?
So at FP16, I alone keep a 1,664 GiB system occupied all the time?
It depends hugely on what "rent usage of this model through one of the many LLM hosting providers" means. If you're asking them to host the model privately then yes, all of that 1.6T of RAM is likely in use holding weights, activations and KV cache by an inference engine that's only getting/answering requests from you alone. When you aren't actively using the model the hosting process is still active and waiting with all of that memory still wired to it.
As background: For the most part VRAM oversubscription/paging/swapping isn't a thing in the same way that RAM for a VM often is. There are some approaches to it, but (to my knowledge) not at that sort of scale.
There are some systemic reasons for this, but very broadly speaking the GPU vendors are building toward the highest bandwidth and lowest latency possible, and the overhead/complexity of something like protected memory modes serves neither of those priorities.
No, a cluster can server multiple users at the same time, providers cap the tok/s so that one cluster can run inference on multiple inputs at the same time. OpenAI with their new ultrafast mode is probably reserving the whole cluster or prioritizing requests of ultrafast users above others with a higher tok/s hence the high price and high speed. There's many other knobs providers tweak that they don't show the users, for example I doubt many providers are hosting the full FP16 version.
The article isn't just about running locally though. The author is saying it's super cheap to run the model through Opencode Go (and presumably OpenRouter etc.) Personally I'm always most excited by models I can actually run locally, but even these huge open source models open up the competitive landscape for companies to let you call models via an API or just lease compute. And they don't have to charge you to offset research, training, huge staffs of the best minds in the world, crazy PR etc. I think that's a big win for customers and buts competitive pressure on the frontier labs as well.
I think, considering the size of this model, it's closer to a Pro than a Flash on everything other than speed
I did somet math and completely gave up on the idea of trying any worthwhile local model and figured I'd rather pay the 15-30 USD per month via subscription and/or API key combos for years than buying a local setup which might go out of date very fast, if it doesn't goes kaput just out of warranty. I won't be surprised if RAM scarcity is an concerted effort to herd people towards the remote models :)
> Even so why would anyone not sleep on a model they cannot run?
Because it's an open model so providers compete on price.
I just un-retire my pair of 1080Ti for some small models development because the current GPU prices literally make me sad.
This “blame sama for memory prices” meme is so tired.
He gave demand signal so many times years ago and was mocked for it and now we have the consequences of industry not taking him seriously.
It's dangerous to go alone. Take this: [1]
I reimplemented most of the features of the Deepseek v4.1 flash paper (apart from quantization aware training which doesn't make sense because my implementation uses float32 precision anyways)
I'm currently learning how to distill reasoning traces (check my other github repositories) but I think that a locally selfhostable deepseek is possible with my mixture of experts sharding mechanism. I decided to optimize everything for CPU parallelization, with the idea that the KV cache and meta model have to run from CPU RAM anyways, so the experts can also be loaded/unloaded at runtime if needbe, to save more RAM.
My assumption is that the KV cache optimizations in combination with the CED and compressed attention features are the reason why v4.1 flash has so few hallucination problems and such a strong self-lookup/thinking behavior. But that's more a gut feeling, need to evaluate and test this more thoroughly.
Anyways, would love to see someone train this on their own datasets. Currently my pipeline is kinda optimized for parquet and zim files.
[1] https://github.com/cookiengineer/gonano
The bulk of its weights are natively MXFP4. And engram values don't need to be in vram.
one can make a fine gaming pc for ~$350
1660 ti, 4790k, 16gb ddr3
You can run it locally for the price of a decent car, or run it (hopefully) privately on somebody else's hardware at vast.ai or a similar provider for much less. What's not to like?
No, you won't get frontier-level intelligence on a 1070Ti. Yes, it should be illegal to do what Altman did. Since we clearly don't live in the best of all possible worlds, we need to settle, and DS4.1 Flash is a good place to do that.
For tasks that don't require vision I personally like the NVFP4 quant of GLM 5.3 from Local Inference Lab better than DS4.1F, but they are both well beyond awesome.
He doesn't ruin the cost of memory. Advances in memory size and speed are now in full speed mode. Expect drastic increase in the upcoming years. Big factories are in the making and planned. Gigalab in the US and many others in the east. Since 2010 we have computers with 16gb as being normal. Finally we are moving into a new era where the standard will be 64gb next year and 128 in 2028. Hopefully we reach 1tb in 2030.
I've read the exact opposite, vendors are reducing the standard from 16GB back to 8GB
That's only temporary till production meets demand again.
Which is not going to happen in the upcoming years
What are you talking about? The model is native NVFP4, why you run it at any precision higher than that?
Why isn't the author worried about sending her/his ideas to DeepSeek online (instead of hosting it and using it locally)?
I have a pretty large, complex project I've been building with heavy AI use (new language + compiler). I was following a 'strong model as orchestrator launching cheap models as implementers' pattern, but I recently trialled just using Deepseek-V4.1-Flash as the model for both layers because of the cost savings (with mimo v2.6 flash on code review agents for some decorrelation).
I was previously using GLM-5.3 as the orchestrator, after switching to DS anecdotally there was an unnacceptable quality loss, mostly around not taking all the relevant context into account when making decisions, pulling new design out of thin air without discussion too often, and being way too wordy and rambly in documentation despite prompting to avoid it. There's a lot of docs, rulings, core concepts, design philosophy to uphold and DS was just not cutting it.
However, it's perfectly capable of being the sole agent for all of my well specced implementation tasks. I've gone back to GLM as the orchestrator.
On the Claude side of things I was previously following "strong model directs weak" with Fable directing Opus/Sonnet (its choice per-task). Since Opus 5.5 came out I've just been having Opus direct Opus.
The sub-agent separation is still valuable to keep context clean for the orchestrator, but I just have no reason to use Sonnet as the grunt-work implementer because I'm finding it hard to run out of tokens with Opus 5.5 on a $200 subscription plan. It's really really good at subjective quality of work per token used.
I would probably go that route if I could use other harnesses with claude models, but I don't want to be locked in to claude code, and their API pricing (which you need to use it with other harnesses) is so much higher than subscription.
Can't wait to just use the hand made Jai and laugh at everything else built with AI :)
I'm paying for the heavily-discounted subscriptions, not the API rates. There isn't really a cost gap for me. DeepSeek doesn't have a subscription to compare to, but when I compared the GLM 5.3 usage I got from a $100/mo Z.ai subscription compared to Opus 5.5 on a $100/mo Claude subscription, there wasn't a big gap. And GLM 5.3 is very clearly not a frontier model (deepseek v4 seemed a lot closer, but I didn't use it enough to really say for my workloads).
I don't think those subscriptions nave negative contribution margins, either. I think we're seeing a lot of price discrimination by the big labs, and huge margins on their frontier models. The fact that they have been cutting prices to their second-biggest tier of models (Opus/Sol).
Open models catching up and collapsing these margins would worry me if I were a shareholder in the big labs, but as a user, I really doubt that the western labs have bigger environmental impact just because they have higher API costs, I think they have a ton of efficiencies they aren't sharing with customers yet because demand is so high.
The tightening of subscription value has already begun. dsv4.1f is already worth paying for at market API prices. Maybe it goes to 2x because apparently no one has figured out how to match DeepSeek's insane caching efficiency, but I don't see it getting much worse than that.
Plus you can also get dsv4.1f subsidized. OpenCode Go gives 4x if I understand their pricing correctly. Anecdotally, I feel like I get way more out of my $10/mo OpenCode Go sub for the price than my $20/mo ChatGPT, even using gpt-6.1-sol high which is very cheap, and I have yet to convince myself dsv4.1f is a worse model.
It's really not cheaper than frontier subscriptions. It's getting closer, and it's a great model, but it is not more value per task than the frontier subscriptions. Don't be swayed by the token costs, it's very chatty, like 3x more tokens for the same task as sol. I used dsf 4.1 full time for about a week.
It blows frontier API pricing out of the water, but again, look at cost per task, not token usage. Still easily wins though for my work.
I do think it's the most viable alternative I've seen so far, and that applies pressure to the frontier models. Should subscription prices hike or become unavailable for some reason, I know what I'll be using.
When pricing this, it's important to consider whether or not you want to opt out of data training. You won't get the advertised rate. Also the dsf 4.1 subscription providers are throttled af... and of course they are, because otherwise they'd be haemorrhaging money.
I am looking at cost per task (and speed per task), both benchmarks and anecdotal experience.
What are you talking about?
DS4.1 Flash not really cheaper than frontier models???
It is insanely cheaper.
Oh man. v4.1-flash has been an sbolute game changer for us. We run all our Personal Assistants now on flash (thinking high) by default and it works incredibly well. There is really no need for basic agentic tasks that might require Kimi K.3 or GLM-5.3 levels.
Once its gets juicier, we let flash launch specialized subagents with specific models. GLM-5.3 for coding or Kimi K.3 for research and critique.
But as a main driver. I love flash. And it brought our bill down by A LOT :D
What personal assistants are you using?
What is the cost of access like for DeepSeek-v4.1-flash, compared to GLM-5.3-flash via ZAI's Coding Plan? Because that's what I use; and often hit the "wait". I wouldn't mind trying a new model subscription or even API access which hits around glm-5.3-flash level weight class (which seem to be enough for me; with quite some human suprvision and nudging) but gives muuuuuuch moooore tokens for the same price.
(And what are the preferred providers?)
>We run all our Personal Assistants now on flash
are you worried about sending all your data to third parties, especially if they're in different countries?
Not shilling for them but Ollama cloud hosts domestically with ZDR afaik. I run 95% of my open weight inference through them. The rest goes through Opencode Go $10 plan (which is enough to run 3 hermes agents on DSF 4.1 and leave plenty of left to experiment with when new models drop).
Just a reminder that any API using a Cloudflare TLS certificate isn't ZDR.
The model engine provider might be ZDR, but the service as a whole isn't.
did not know. ty! have my updoot as thanks
I use it through OpenRouter, which has ZDR enforcement.
I have asked OP that question but I think there are providers who are not in China and they just host the model/inference.
Just use a provider hosting it in your country especially if your country has major data centers then its the same as using Anthropic or GPT of GCP Model Garden or AWS Bedrock
nobody here is talking about running frontier level intelligence locally so if you’re Chinaphobic and prefer layers of corporations siphoning your data in between you and the party there are plenty of options instead of directly to the party
Have you compared it against actual SOTA models like latest Fable or Astra?
The author explains this very well tbh:
> Today's models are now good enough for high-quality unattended tasks. Chasing the latest and greatest is silly. It is fun to see the new Fable capabilities, but the tasks we throw at them are usually ridiculous (maybe even insulting) if you believe in LLM sentience. It's like asking a math PhD to organize the files on your desktop.
I'm using DS V4.1 Flash as my main model since their release and it works great for all my coding tasks. My setup is OpenCode Go subscription and obra/superpowers skill.
The only times I try to change models are on general planning tasks (like research this codebase for tech debt mitigation opportunities) or if I need deep research which would benefit from searching the web, in which I still think Gemini is still the best because of the speed and access to google search index. But these are not even 20% of my daily tasks.
Of course there's still a huge performance gap
but DS 4.1 Flash is good enough for most tasks
I have been using DeepSeek 4.1 flash intensively for over a month. If I run it all day long it costs $1-2. Its fast. Previously I was always quickly running up to my Claude/Codex 5 hour window (on the $20/month plan). The cost savings of DeepSeek is real as shown in this article and I am using subsidized plans.
DeepSeek is horrible at grilling sessions (the /grill* skills to make technical decisions). It doesn't know how to explain things. Maybe the skill could be adjusted. It also doesn't come up with as good solutions as Opus/Sol.
What I use it for is
Previously I planned with Opus/Sol/Astra and then I used DeepSeek for coding, and then reviewed with Opus/Sol/Astra. With the cost improvements to Opus/Sol I am trying to use them for coding instead now so there will be less back and forth review needed.They are all working together in Pi using the extension @tintinweb/pi-subagents where my workflow skill is calling different subagents that use different models.
Luna is cost competitive, but doesn't score as well on intelligence. I do need the intelligence for most of what I use it for, so I am not motivated to use Luna. Haiku also doesn't seem like a competitive price/performance mix.
Luna is 1 point being on AA's index at 1/4 the cost, yes it "doesn't score as well" but paying 4x for 1 point is crazy if you're going off benchmarks.
AA has Haiku 5.5 as cheaper than 4.1 Flash (both on Max, which isn't ideal but what can ya do) and a 4 point intelligence gap.
Why do people like to think open models are more competitive than they are?
It is super bad on a bit more complex workflows and starts repeating same errors with the same tool until the cycle breaker hits.
6 is worse than 5.6 here.
But it is amazing on generating a report on content generated by better agentic models such as DeepSeek or GLM, which both do a mediocre/bad job on reports.
DeepSeek's own paper advises against using Max, showing that it normally doesn't perform that much better. I am not using it on Max, so that's not a useful benchmark for me. I have seen other benchmarks where Flash does significantly (30%) better than Luna.
Because DeepSeek is not "a month or two" behind as claimed in the article.
These open models still did not beat February's Mythos / Fable 5.
DeepSeek 4.1 Flash is behind GPT 5.6 Sol, and that one is left in the dust by the excellent Opus 5.5.
Rumors say Anthropic is holding in reserve the big improvement, Fable 5.5, for the IPO.
It's plausible that open models are 6 - 12 months behind, and there is no "good enough". As long as progress doesn't slow down, leading labs have nothing to fear.
I was thinking about this earlier today and I came to the following question:
If you had a model 10x as capable as the best model out today, but it cost 100x more, would there be a market, and, if so, how big?
I think there would be a market and I think it would be large.
So, I agree.
Many people would, and you'll find that they're building crappy webapps where you dont need SoTA. Like seriously who needs these frontier models?
Unless you're doing some extermely difficult post-grad lvl research, you do not need a 100x PhD research assistant, especially not for whatever silly SaaS product most people are building.
There's people at my job that get so much more done than everyone else using Fable. and all they use is the fastest cheapest models. I'd say the people who are using sota models for everything are doing it just because they prefer to be lazy.
You simply do not need these frontier models, they outgrew most people's needs 6 months ago, but for some reason people still want to run a 700k rack of gpus full throttle to center a div for them.
Doing what? How many jobs involve solving Millennium Prize math challenges?
99% of everything is CRUD LoB apps.
I am coding CRUD apps with a mix of astra, sol 6.1, fable and opus 5.5. A more capable model would still benefit me imo. Being able to follow high level guidance better, and being able to harness other models for each task would be a big improvement.
that just sounds like openai/anthropic cope/propaganda, based on absolutely nothing objective lol
even their harnesses are far surpassed by pi and opencode at this point
also sick 'rumors' lmao, apparently marketing through rumors is in vogue these days
>> that just sounds like openai/anthropic cope/propaganda, based on absolutely nothing objective lol
Nah. There are benchmarks. They are free to look at. And they paint a very clear picture.
There absolutely is "good enough" and I agree with this author: DeepSeek 4.1 Flash is plenty good enough for all the things I would trust an AI to do at my job.
I am 4.1 maxxing on commandcode GOAT Plan + api rates with oh my pi for the last 4 weeks, it's absolutely amazing and crazy fast, it's alright if it makes a mistake, I have enough time to iterate again, I have also added an advisor layer of mimo 2.6 pro which does make it a notch smarter. Getting haiku 5.5/sonnet5.5 to work on plans and letting 4.1 flash work through it is helping a ton too.
I am a big ChatGPT fan, all our team has ChatGPT Subs, but the TPS across all models including luna is just so damn slow.
Commandcode giving 60$ worth of Deepseek for 10$ is just genuinely goat.
And it never says no for cyber tasks so that's a big win
Yeah. I've been enjoying Coralbricks 250-350 tok/s speeds and it is hard to go back to slower models.
Lithos promises even faster speeds if you want to pay more.
I had the same experience using ds 4.1 last couple of weeks. It’s insanely good for the price. I’m doing mostly web dev it excels at everything I throw at it. The pricing is ridiculous. I canceled my gpt subscription and haven’t looked back hope the pricing stays like that. I almost never need a better model. I still keep my Claude 20$ sub for now but I feel like one more iteration and I won’t need even that anymore I hardly use it
If DS4.1 impresses you I would be really interested to see your comparison to GLM 5.3. I switched from the one to the other and even if GLM 5.3 is a bit slower I don't think I'll be going back.
DS4 (not 4.1) crossed my dont-care threshold and I genuinely stopped paying attention to new models. I'd love to try GLM 5.3 but I just don't see any point in spending the effort any more. I can get passable intelligence for a bargain price either direct from China or from a ZDR EU provider for a small markup. Paying 10x more will not make me 10x happier, it's unlikely to make me even 1.1x happier now I've got some intuition for the natural limits of these models.
I don't even bother checking how much I spent on API any more, its well under $30 over the past 2 months despite daily constant use. Who even needs a subscription at these numbers?
GLM 5.3 is very impressive and definitely better, but it also at least 4x the price.
On that note I’ve been subbing in MiMo-2.6-pro when cost is an issue, which is super cheap and also performing really well.
IDK what happened today but I used GLM-5.3 as usual from Ollama cloud and it was so fast it generated entire documents like instantly.
The reasoning and the result document were done after less than 1 or 2 seconds.
Have Ollama suddenly bought GPU capacity?
There is also GLM 5.3 Flash
What inference provider are you using?
Been using Flash 4.1 via the ante harness to blast through a GBA recomp. The ante team has pushed hard to make Flash 4.1 perform well under it. So far, I've maybe spent $10 over the last 3 days. Its a real workhorse and works much better in this harness
https://tangled.org/astrra.space/ds4-recipe is an incredibly cool writeup on making deepseek v4.1 flash run really fast.
I did a test on this a couple weeks ago. What I found was that the chinese models were far better than API rates, but about comparable on price vs the subsidized subcription model (chatgpt). Also it was my experience that codex completed tasks quicker.
That said, it's my best understanding that these american companies aren't profitable and will eventually raise rates (the old uber trick) so I'm keeping myself ready to switch when that day comes.
The question seems rhetorical but I think there are two reasons in some combination. First is it there is some awareness lag here. That lag can be on the producer and consumer side. Software enterprises are pretty slow to adopt new things and slow to try new things so they might only be aware of openai and Claude as options. Plus there are some scariness because deep seek is a Chinese model and therefore export restricted - never mind that there are American in European providers.
The other reason is more interesting. Maybe the frontier providers think that price performance is irrelevant in light of very powerful frontier models that can start the RSI loop and or a huge displacement of work and a winner take all economic situation. After all if frontier providers earn everyone's money then you won't have any money to spend on any model 100x cheaper or not.
I think it also has a bit to do with the AI sector of tech still moving at lightning speed.
Theres already models that outdo DS 4.1 flash in cost/performance. Luna 6 on max effort for example. Luna also doesn't care what time of the day it is for cost calculation.
And I'm sure by the time people ask why Luna 6 is being slept on there will be another cost/performance king
Luna is very slow and bad at agentic tasks. DS runs circles around it and there are US providers providing cheaper rates no matter the time of the day.
While using DeepSeek v4.1 Flash I was architecting a system and I made a mistake of drawing the RPC boundaries at a wrong place that did cost me in so many ways.
I realized that mistake and guided DeepSeek where it should be.
Next I fired Fabble 5.5 set to high to check if the hype is real about Fabble. It exhausted 89% of quota and came up with NOTHING that DeepSeek hadn't flagged itself already in its notes.
Do you mean Fable 5.1? Or Opus 5.5? I'm not sure what you're working on but for me DS 4.1 flash isn't nearly at their level. For the price it's obvious very impressive, though Luna 6.0 is excellent too.
The problem with benchmarks and proprietary models is that one day a model is best at doing X, another day that's not so sure. And anyway, we are not throwing the same X.
I've found supposedly smaller and, less performant models do better on certain tasks. I end up using several models, sticking to what my unconscious statistical observations tell me to use for the kind of task at hand.
care to share what exactly are you working on?
Fabble 5.1.
> With my OpenCode Go sub of $10/month, DeepSeek is basically unlimited
My OpenCode Go monthly window was scheduled to reset this morning. It was sitting at 22% used despite me using DeepSeek V4.1 Flash heavily as my implementation agent the past couple weeks (I use gpt-6.1-sol high for planning/orchestration).
I had 1.5 hours left so I fired up first 10, then 20, and finally 50 concurrent subagents all working on reverse engineering C code from an old PC game. They found over 100 new functions.
This is the first workload I've found that could make a dent in my sub. It got my 5 hour window to 85% used, but sadly my monthly was still only at about 35% when it reset. So that cost maybe $2.
How are you guys doing orchestration? I have been fumbling around in my free time trying to build something for myself but is there a repo or something that just works?
I might not be the best person to ask. I use Pi harness in tmux and just ask my current agent to spawn interactive pi instances in new tmux windows, create a sentinel file for each of them, and monitor the sentinel files for signals every 2 seconds.
Currently have auto compaction turned off. When the orchestrator's context is getting close to full, I have it write a handoff markdown file and point a fresh agent at it.
I do feel like I'm getting close to the point where I might be ready for something more sophisticated, especially wrt to subagents communicating with the orchestrator.
Many orchestrator harnesses exist.
Check: https://agentmgmt.dev/ and find the one that works for you.
I quite like Paseo (been maining it for a week), but Orca also looks good.
I think the interesting provider to cross-check this assumption with here is Meta, who is clearly freaking out, and is currently providing Muse 1.3 even cheaper so long as you are willing to share data with them
Whatever the question, Meta is the wrong answer.
> so long as you are willing to share data with them
I don't think so.
This model finally got me off my Claude Max subscription. I’ve found it superior to Opus 5.5 in certain use cases, and certainly faster.
I’m convinced that I’ll have good enough inference on my laptop at reasonable speeds within the next year.
I think because we're all just using it thinking we have found the "model for me" and never mentioning it to anyone because what would we say? It's good. It's a bit like the Logitech MX Master, as more and more people assumed they had found the ideal mouse for their purposes, it quietly became the professional standard through sheer adoption.
What about MiMo v2.6 Pro? It’s throughput is slower by default (UltraSpeed is faster than DS4.1F) but is above the pareto line, and cheaper.
see https://artificialanalysis.ai/models/releases/comparisons?co...
Technically yes, but has been reported to be quite benchmaxxed. In practice Deepseek Flash 4.1 and GLM 5.3 might therefore still outperform Mimo 2.6 pro.
I’ve been using it a lot and it’s performing really well. Not GLM 5.3 levels but it beats Deepseek for my use. I’ve used it on long running coding tasks, though mostly prototyping, but it’s done a great job at very low cost.
I'm using it quite extensively in my PlayStation decompilation harness
[delayed]
Enterprise is not freaking out because DeepSeek 4.1 Flash does not actually occupy a spot on the Pareto frontier for non-coding enterprise workflows. We see this at my employer, focused on non-technical knowledge work. Luna 6 and now Haiku 5.5 are both very competitive if not better on all axes that we care about
Deepseek V4.1 Flash is a hidden gem, really. Not to mention, you can easily get it through many providers that offer zero data retention and consistent speeds above 200 tokens per second!
It's a solid little model, and I appreciate DeepSeek's commitment to the bit in releasing a brand new pretrain, double the size, numerous architectural innovations as a ".1" release over the excellent DeepSeek V4 Flash.
DeepSeek isn't even on my mind. I use the frontier models and can get the best in the industry for a relatively cheap price.
Yeah. $100 for Claude just about gives me all the usage I want, as a more or less full-time hobbyist having it work in the background most of the day. I was trying to economize by having a local LLM, then Deepseek, then Cursor/Grok, and then I got a taste of Opus 5.5 and I simply cannot go back to having to carefully spec things out and double-check work. I just let it decide, Opus or Sonnet for the next task, and I get almost perfect results. Probably similar with OpenAI's models.
The token-equivalent monthly spend is > $5K+. If Deepseek's token cost is 20x cheaper, that's $250/mo, and I'd be spending a lot more of my brainpower babysitting it and getting worse results.
For business/team accounts that pay per-token, maybe I can see the "freaking out" being warranted on the part of the fronter labs. But as long as they're willing to subsidize their end-user subscriptions, I'm not going to move off of them until the alternatives are truly at their level.
[delayed]
I have subscriptions to OpenAI and Claude but use DeepSeek 4.1 Flash for my coding agents.
It costs pennies and you got really great output.
The author is spot on.
Because in two weeks someone will be asking why I'm not freaking out about AlphaDolphins 0.3 Zip and then in a month FrozenMonkey 2.5 Artic.
Probably because Luna is faster, cheaper, and approximately as smart
What would freaking out look like, or is this just a stupid bloggish title flourish?
Is OpenAI coming in $20B under a sign of "freaking out"?
Agree that people leaving the big two companies is going to be hard to keep a pulse on prior to IPO.
Anthropic and OpenAi are in the news, so they get the press and people go and try out their product. Large enterprise businesses are going to make larger, longer-term contracts with them and are only going to pivot if they think switching costs are easy or if they think the provider won't deliver.
The other inference producers are less well known or you need to get your cloud sales rep to tell you how to switch to them as a provider rather than Anthropic or OpenAI.
I use OpenRouter, I know switching is easy, but larger businesses tend to work in yearly cycles. DeepSeek v4 Flash came out in late April.
I agree OpenAI and Anthropic are going to struggle when the median price of running a smart-enough model keeps falling.
Edit: I also think demand for hardware will be rapidly absorbed by other companies if Anthropic or OpenAI stumble. We've finally turned hardware directly into runnable intelligence and people are not going to go back to the old ways.
It would look like major chaos in the markets.
People tend to conflate the question "is AI a useful technology?" with "are the AI companies going to do well?" but they're surprisingly separated in practice, with either one able to be true while the other is false. There is a lot of money tied up in a lot of hardware with a lot of loans made against that hardware as collateral all based on the assumption that AIs are going to need more and more and more and more hardware and whoever has the hardware wins. If a much better model comes out that requires vastly less hardware, or even more accurately, merely charges vastly less than the current AI companies, then to a first approximation (barring Jevon's paradox, and bearing in mind there's no timeline guarantee on that) all that hardware becomes much less valuable for being grotesquely oversupplied relative to what is necessary, and even though that would generally make AI objectively more useful than it was before, it would cause mass financial chaos in the markets.
The markets need a very particular rate of progress. It isn't entirely clear to me that it's even a possible rate of progress, it may be overconstrained, but they certainly don't have plans for the AI models to get commoditized on the timeframes of these vast, vast array of loans being made against hardware as collateral. Spend a metric shit ton of money to kill all your competition then charge monopoly rent on the one thing absolutely everyone needs doesn't work if you can't economically "kill all your competition" because the economics favor them in the spending spree.
And then, based on the fact that this is not even remotely complicated logic, there are plenty of people who are fully aware that they have a lot of money tied up in not running around telling everyone how wonderful the cheap models have become.
It's also unclear whether those who approved those loans understand GPUs depreciation. In any case, progress in software but also hardware could bring chaos and ruin their house of cards.
> It's also unclear whether those who approved those loans understand GPUs depreciation.
This also assumes heavy utilization, though. If there's heavy utilization, it might mean they're doing well. If they're all spinning, it's time to raise prices.
Only if utilization isn't at a loss. Right?
they should be freaking out because every time the chinese labs or non "frontier" labs release a model that is only a few months behind and much cheaper than the openai/anthropic models it shows that they don't deserve their valuations
Perhaps, OR it might be that most people in the markets (think that they) are not all that exposed to the valuation AI labs and so their eventual collapse doesn't matter.
Or perhaps they consider the upside from cheap Chinese models to hedge the effect that OpenAI/Anthropic collapsing would have on their portfolios. This would make sense for (hedge funds holding) most companies: they don't really care about who supplies the AI, as long as they get it at roughly the same price as their competitors.
ironically, at my large enterprise, they aren't yet distinguishing between "chinese models" and "chinese models hosted at microsoft foundry". so far it's just _banned_. i'm not at all pretending it's like that at other orgs.
They might be. They would delay public admission as long as possible, because public admission would make stocks go down.
I have used over 40B tokens and spent over $800 on DeepSeek API over the past 30 days, mostly on V4.1 Flash.
It's good, and you can do most work with this. For complex software implementation you need to split your runs into various phases, build in verification, and use subagents so that work gets another audit and repair pass from the lead agent. You can do pretty much everything then. Frontier models can do without compelx workflows, that's the difference.
My anecdotal experience is that I can’t even trust DS4Pro let alone Flash. I always have to have Sol reviewing the code.
I built mjolnir in large part so I could have Opus manage DeepSeek Flash subagents. It's phenomenal and extremely light on the Claude tokens. https://github.com/BrokkAi/mjolnir/
And yes, Opus is enough smarter than DSF that it's worth the extra steps. This ranking is from live tickets, no contamination: https://slopcop.com/power-ranking
Probably off topic but this is pretty wild to bury in the readme
> By default Mjolnir sends recent prompt and reply text and help-search text to TypeSafe's hosted Jev classifier through a public proxy
Because GLM 5.3 Flash is even cheaper?
Opencode Go gives only 6300 requests for glm and 23 000 for deepseek. And, if I wanted to, I would be able to do all my work on $10 plan with deepseek. It’s very cheap.
DSH is listed as a https://opencode.ai/docs/go/#known-problematic-clients :)
> and 23 000 for deepseek
How did you calculate it? Based on per 5 hours max request allowance?
Nah fam, not true, also DS edges it out on coding / dev tasks.
That is opposite to my experience so far, can you describe your coding tasks? Mine are systems level code, utilities, operating system code, networking and real time control stuff.
I also prefer GLM-5.3-flash to DS-4.1-flash, but it's close. Since Z.ai has been offering essentially free GLM-5.3-flash tokens on their coding plan between 8am-6pm pdt I've been using it a lot... though that ends on Oct 10 IIRC.
On design tasks too, for me.
Don't think so.
here you go: https://artificialanalysis.ai/models/releases/comparisons?co...
DeepSeek 4.1 Flash 0910 is perfect for M5 Ultra 256GiB. Running it fully resident in RAM, prefill at ~2500 tok/s and decode at ~40 tok/s. Probably tons of room to improve from there.
Which quantization are you using there?
You are running this now? :o
I like GLM 5.3 Flash, it seems good enough for coding features if you have a good structure and a good AGENTS.md
Don't you run into it sometimes outputting a few Chinese characters, or Cyrillic, for no apparent reason? I fear it writing some nonsense in the code or the terminal. DeepSeek V4.1 Flash doesn't seem to do that.
That hasn't happened with GLM 5.3 yet but with DS 4.1 Flash it did happen and it also had a tendency to loop.
To be fair even OpenAI's and, to a lesser extent, Anthropic's models do that sometimes
There was a moment 3 months back where the sentiment was that cheaper models were the way to go. Since then the pendulum has swung back.
What blows me away about this model is it's speed.
It's way faster than Opus or any of the GPT models.
I have a coding harness which is opencode plus a few skills relevant to my workflow. Deepseek 4.1 Flash does very well in this environment. I haven't noticed much difference quality wise compared to Opus 5, which I use in my day job as my employer pays for it (although I'm considering using DeepSeek here too given how cheap it is).
Because subscription plans are cheaper, only enterprises paying per tok pricing should be freaking out
For non-coding tasks it may be fine. But for coding, Opus 5.5 is just a completely another level than something like Deepseek 4.1 Flash.
Opus 5.5: TIME 9.3m COST / $1.99 / SCORE 99/100 https://jonclegg.github.io/pacman-bakeoff/#claude-opus-5-5
Deepseek 4.1 Flash: TIME 2.8m / COST $1.89 / SCORE 72/100 https://jonclegg.github.io/pacman-bakeoff/dev/#deepseek-v4.1...
Downvoted for facts. What is this Reddit?
Everyone here is so unhinged.
I was just using a bunch of models in Cursor to review a project. I went looking for DeepSeek and it was not one of the options.
Would be cool if they added it.
Since they adhere to the same API spec, you can hook any model. It takes one line edit in /etc/hosts
There are some quirks if your harness use unsupported features of course.
I don't know about "freaking out", but I'd say I'm having a good time here with DS 4.1 flash.
America bans Chinese models sooner or later just the China banned American big tech
Is it known why unsloth seems to not have touched DeepSeek 4.1 Flash?
It's still lacking llama.cpp support, and the work on that isn't moving very fast either. Looks more like a general community issue, where this model isn't drawing much interest.
You can ask them directly, Daniel Han-Chen is pretty responsive.
People have been raving since forever about Deepseek, but if one looks at the CoT, it's evident that it's way way stupider than frontier models (there's a reason why it's cheap). It's laughable to compare Deepseek 4.1 with Opus 5.5.
I've benchmarked, rigorously, deepseek-v4-flash for programming and personal use, and it is definitely less smart than Qwen3.8-flash-next (which in turn, is not terribly smart).
Local models are also really slow, unless one spends insane amounts of money.
Having said that, Qwen3.8-flash-next is an impressive evolution; it reaches the small versions of the frontier models (like Sonnet) - but again, it's massively slower and not 100% reliable (including: stability).
The argument that most people are making isn't that dsv4.1f is better than frontier, but that it's good enough for most tasks, faster, and way cheaper.
> if one looks at the CoT, it's evident that it's way way stupider than frontier models
Frontier models don't show the full CoT
The COT isn't an end all be all. Research has shown that the COT isn't necessarily what the model is actually thinking.
Honestly for me the intelligence gap between DS 4.1 Flash and Muse Spark 1.3 makes Muse more worth it for me, especially on a $10 OpenCode Go sub, with the caveat that everything I use it on is open source which makes the fact that I'm sharing it with Meta a little moot because it's already published permissively on GitHub anyways.
> shrank the KV cache by roughly 437X
Can't you just say "shrank to 1/437th the size"? It's not that hard.
I had it make 25 different things today and it cost $0.70
It’s disgustingly good value. I find it capable of doing anything I want.
Obviously can’t use it at work, but for home projects it’s awesome.
> Obviously can’t use it at work
I do wonder how long it'll be before a us-hosted offering is available via bedrock, copilot, etc.
There are already US hosted offerings on companies like fireworks.ai.
I would love to get them more better, it's good not a bad thing.
Because engineers are lusting for 1000 tokens per second. You can only achieve something like that with OpenAI.
I'm no expert, but it think that it's the pricing on GPT-6 Luna. I'm also guessing that it's been underpriced just for this reason. I also don't think it's all that great, but it's definitely very cheap.
If it's underpriced, it's a loss leader to sell the other models, so it actually can't be too good.
I really put these things through their paces because I use them to review and work with new abstract game rules and models, so they're always flying blind. Luna misses the obvious (and more importantly, the clearly explained) consistently. My second prompt is listing all of the points in its first response, and saying "No, it doesn't work like that." The third prompt is picking out the two or three suggestions it made after correcting itself on all of the original points and saying "That's how it already works." The fourth prompt is "Now that we're done going over the rules, can we start?"
I actually feel like 5.6 Luna seemed better.
In my experience it just takes so much longer to arrive at "done" state for me. It thinks for soooooo long. I guess if you're running 12 sessions at once you don't really notice.
No they can't.
And as long as I pay as little for claude opus 5.5 i do right now, i'm using it.
But yes i'm glad that we have alternatives.
i thought 6.1sol copied the caching architecture so this isn't such a big deal no more
because it doesn't work very well?
if you have a legitimate coding application, it isn't very good. if you have some kind of inauthentic activity, which could be what it is trained for for all sorts of reasons...
What is your evidence? Deepseek v4.1 Flash is by far the most popular coding model on openrouter, having processed 38.7T tokens in just the last 7 days, over 3x the usage of the 2nd rank model.
So I ask again, what are you basing your assertion on?
my own usage of deekseep v4.1 flash, and that among the dozens of great programmers i know, not a single person is using it
BUT. they are employed to do / deciding-to-do authentic (if often meaningless) stuff.
here's a short list of inauthentic activity that claude and openai refuse to do:
- chat services that, when you ask them, say they are not chatbots when they are
- code to work around software licenses or DRM
- code to scrape or download copyrighted material
- directly cheating on homework
- adopting a persona in social media that spreads misinformation or propaganda
this is but a short list. but ask me, "are there enough inauthentic activity demands such that someone who CANNOT USE claude or gpt as the LLM would use dsv4.1 on openrouter instead?" yes. i mean there are whole countries right now where the culture can be summarized as, "bottom to top, inauthentic activity." i am surprised it is not more usage!
So your argument is that the 38T of tokens used in the last 7 days is by moron programmers or people doing "inauthentic" tasks? You think the person who made this post is also an idiot?
Do you realize how incredibly delusional/self-centered you sound?
do YOU know anyone gainfully employed in programming who is using dsv4.1 to do work? what kind of work is it? why don't you ask them if it is good?
i don't know why you have to attack me personally, i think you're a bright and otherwise nice person and you understand the thrust of my POV.
Why would we freak out? The systems we use have always gotten better, faster, cheaper with time
AA shows Luna at 1/4 the price, 1 point behind on intelligence matrix with a 38.
Haiku 5.5 is 23% cheaper with a 4 point intelligence lead.
I'm on subscription usage so I can't compare Flash 4.1 to them directly but the OP has his head up his ass if he thinks Opus 5.5 is the best point of comparison. Why is anyone using Opus if the new Haiku is indistinguishable /s
Just absolutely terrible post, admits to using Opus for review but claims its intelligence isn't needed, why aren't you using Haiku or Sonnet then?
because we've all figured out that AI is just a huge grift?
Not comparing to gpt-6-luna which seems comparable and priced well.
You might also choose to pay money for a service that provides real value instead of actively choosing to support the Chinese deliberate effort to undermine this country.
Your comment seems to imply there's a good guy in a this. I just see the inevitable end of an era, championed by predictably selfish actors.
Oh, the writing is on the wall. Wait till you hear that European “sovereign AI” is just running GLM 5.3.