So OpenAI cut Luna's price by 5x, DeepSeek increased price by 5x!
If I'm reading the benchmarks right, they now went from being much cheaper than Luna (but twice as slow), to being roughly same price (but twice as slow).
So all else being equal, where I would previously have used DeepSeek, I can just use Luna, and get the same result twice as fast?
(Yeah I know benchmarks are mostly nonsense, but the ones measuring time are real, and it's the most precious resource.)
Old DeepSeek Flash 0731 prices have been independently reproduced.[1] The issue is DeepSeek being inundated and not having capacity to serve the demand, hence the price increases to significantly dampen demand. Never mind international demand either--just think about the magnitude of Chinese domestic demand. Prices for anything related to AI or computing in general (mobile phones, cloud data centre hosting, etc) will continue to climb fast as demand for computer chips _far_ exceeds supply. DeepSeek doesn't have an option other than to just work away on improving their technology in the period of time before computer chips once again become a commodity. For example, DeepSeek's cache ratio for their models apparently leads to 1/2 GPU time requirement versus the second best provider.[2]
Let's say DeepSeek is being forced to use the CANN stack, and the new pricing reflects the cost when 100% of inference is done with Huawei chips. Then, I suppose we can infer that:
* CANN stack is 1.5x~2.3x less efficient in compute
* CANN stack has 6x lower inter-connect capacity
> computer chips once again become a commodity
Ascend 950 is going for $7k to $9k with mediocre looking specs. $16k for RTX Pro 6000, $6k for RTX Pro 5000. This is not looking good.
I pay for Google AI Pro (Bought a year in advance) and Gemini is so bad, I burned through 75% of my five hour allowance trying to get it to fix something.
I pasted the same prompt into OpenCode, set to Deepseek v4 flash free and did it first try.
I'm was going to purchase Opencode GO to try it, but seems my timing is really bad :( hope it doesn't go up too much in Opencode or they find other providers. Bad timing!
As well as the headline in/out changes, people heavily using agentic coding tools will want to note the 6x (off peak) and 12x (peak) increase to cache hit pricing on Pro (since cache hit can easily make up 90%+ of input on long sessions).
DeepSeek was hugely underpricing cache hit pricing before and even after this increase they're still cheaper on that metric than every other provider I'm aware of, but it will put an end to those "I used 1 billion tokens and spent $4" reports.
The problem with DS Flash/Pro is that they are extreme reasoning heavy and step heavy. Step = cache hit. Reasoning = output hit. So the impact on those price increases will be felt much stronger.
I think that Flash is still a usable model but Pro is DOA... Even before the price difference between Flash and Pro, vs the intelligence / problem solving / tool calling did not make sense. But now that gap has widen even more. And there are just too many competitors models now close to that Pro price range.
Especially when we compare that competitive models offer subscription services that easily cut down the token price by 1:10. That makes Pro especially a bad value.
We shall see what the 3th party market is going to do, but i suspect that prices will be increased. If the argument was that DeepSeek increases price as they lack capacity, a company with access to billions, other 3th party providers that need to rent and have less optimized infrastructures will increase prices. Especially if they get hit hard with people moving around.
Its like we always see the same issue with popular models.
* GLM 5.2 is good, capacity issues, API price up, subscription heavy nerfs.
* Kimi K3 is good, capacity issues, API price up, subscription heavy nerfs.
* DeepSeek V4 GA is good, capacity issues, API price up
* OpenAI GLM 5m, 10m active users. Subscription usage is sneakily tightened more and more.
* Anthropic Opus too popular, ...
That is the main issue. The AI users are people who actively easily move between companies. Pushing peak loads to each unprepared company, releasing load on the "less desired". And round we go ...
How could that possibly work? Deepseek was undercutting every other provider by an order of magnitude on cached tokens.
Do they just set a super low caching time and hope that drops effective cache rates low enough? Do all other providers somehow overcharge by that much? Are they just going to sell it as a loss leader?
> Do all other providers somehow overcharge by that much?
This, I think. Cached inputs have an opportunity cost (keeping the KV cache until use) but a hit is basically free. “Basically” - if the cache is offloaded to system RAM or NVMe there’s some scheduling overhead.
From a consumer viewpoint a more interesting metric than the raw costs is
> if OpenRouter is blindly dispatching your requests
This can somewhat be the case, depending on your config. I updated mine to make DeepSeek high priority because I was having a lot of cache misses and reliability issues with the default (cheapest (at face value)) providers, and cost was actually higher overall than anticipated. Was smooth sailing from then; might have to tweak things again now pricing has changed though.
System RAM and/or NVMe storage still has a real cost. And swapping out the context between VRAM and system RAM / NVMe still consumes bandwidth.
I don't have a clue on what the real cost to inference providers comes out to, but it seems really weird that there would be such a big gap, in what should be a pretty competitive market.
That's about 10% over max model len, but to make the numbers round we can call 8GB the upper bound used by the cache for a single sequence. The actual average sequence length is going to be lower, but providers might be using a bf16 KV cache instead of an fp8, so 8GB is probably a reasonable placeholder.
A x16 PCIe 5 link has a theoretical per-direction per-link bandwidth of ~64GB/s, so if you've got 8 cards running you should have 8*64GB/s = 512GB/s total throughput from the root complex (some topology restrictions apply). The math works out such that we can transfer the KV cache for 64 sequences per second.
But here's the thing: we don't need to transfer the KV cache for each token. The KV cache works for the entire sequence, and we're generating thousands of tokens (thinking lol) for that sequence for each request. So the KV cache transfer overhead gets amortized over those thousands of tokens (which takes several seconds to generate, even at 100 tg/s).
The actual numbers need e.g. concurrency and such to plug in, but I think the bandwidth concerns are fairly negligible.
CXL might save us. All the world's old DDR4 to the rescue. Either per box, where the job has to route back, or network attached, where there's now a pool of absurdly fast temp storage.
Why not? They probably only offered those deals because of DeepSeek aggressive pricing. Now that DeepSeek is 3x more expensive it's time to revert those discounts.
It doesn't make any sense unless they are going to exit from inference market. They will be literally one of the costliest option (by output, for flash) if use openrouter as source.
This now places deepseek flash v4 from DeepSeek themselves at higher prices than openrouter (depending on caching). Will be interesting to see if third party prices remain the same.
It does seem to me that DeepSeek themselves aren't so much interested in being a service provider. They do it, and offer the service, but their pronouncements seem to be that they're more interested in being for now closer to a research lab with a longer term play for something more dramatic later.
Unfortunately in this case that's not true at all, there was no provider with genuinely close or the same effective prices (mostly based on cache hit cost) to old v4 flash or v4 pro. People have this misconception that other providers must be much cheaper than the official one in case of open weight models.
If you check on OpenRouter, some other providers serve V4 Flash at seemingly cheaper normal input/output tokens rates, but with a huge caveat: they have at least a 5x increase of the cache hit cost of the official API, some have a 10x+. No provider comes close to Deepseek's old low cache prices, and cache is 90%+ of what matters in agentic sessions.
Closest comparison:
- Deepseek: $0.14/$0.28 with $0.0028 cache hit cost for official API
- DeepInfra: $0.08/$0.18 (cheaper base rates!) with $0.016 cache hit (almost 6x!! Deepseek's current cache cost)
Another great example is Kimi K3, official API is $3/$15 and the cheapest provider on OpenRouter is $2.8/$14, only a tiny difference.
So OpenAI cut Luna's price by 5x, DeepSeek increased price by 5x!
If I'm reading the benchmarks right, they now went from being much cheaper than Luna (but twice as slow), to being roughly same price (but twice as slow).
So all else being equal, where I would previously have used DeepSeek, I can just use Luna, and get the same result twice as fast?
(Yeah I know benchmarks are mostly nonsense, but the ones measuring time are real, and it's the most precious resource.)
Old DeepSeek Flash 0731 prices have been independently reproduced.[1] The issue is DeepSeek being inundated and not having capacity to serve the demand, hence the price increases to significantly dampen demand. Never mind international demand either--just think about the magnitude of Chinese domestic demand. Prices for anything related to AI or computing in general (mobile phones, cloud data centre hosting, etc) will continue to climb fast as demand for computer chips _far_ exceeds supply. DeepSeek doesn't have an option other than to just work away on improving their technology in the period of time before computer chips once again become a commodity. For example, DeepSeek's cache ratio for their models apparently leads to 1/2 GPU time requirement versus the second best provider.[2]
[1] https://nitter.net/thdxr/status/2085377844515922210#m
[2] https://nitter.net/thdxr/status/2087610161636471289#m
Reproduced on the CUDA stack right?
Let's say DeepSeek is being forced to use the CANN stack, and the new pricing reflects the cost when 100% of inference is done with Huawei chips. Then, I suppose we can infer that:
* CANN stack is 1.5x~2.3x less efficient in compute
* CANN stack has 6x lower inter-connect capacity
> computer chips once again become a commodity
Ascend 950 is going for $7k to $9k with mediocre looking specs. $16k for RTX Pro 6000, $6k for RTX Pro 5000. This is not looking good.
According to their post:
gpt-5.6-luna: $0.20 / $1.20 / $0.02 / $0.25 (In / Out / Cache Read / Cache Write)EDIT: formatting
EDIT2: giving up on the formatting :-/
> EDIT: formatting
Keep at it, I believe in you.
My apologies to any mobile users, but for the desktop folk:
TIL: formatting tables on HN is computationally impossible. lol
p.s. Thanks DSv4-Flash, for your hard work of converting a messy table into plain text.
I pay for Google AI Pro (Bought a year in advance) and Gemini is so bad, I burned through 75% of my five hour allowance trying to get it to fix something.
I pasted the same prompt into OpenCode, set to Deepseek v4 flash free and did it first try.
I'm was going to purchase Opencode GO to try it, but seems my timing is really bad :( hope it doesn't go up too much in Opencode or they find other providers. Bad timing!
As well as the headline in/out changes, people heavily using agentic coding tools will want to note the 6x (off peak) and 12x (peak) increase to cache hit pricing on Pro (since cache hit can easily make up 90%+ of input on long sessions).
DeepSeek was hugely underpricing cache hit pricing before and even after this increase they're still cheaper on that metric than every other provider I'm aware of, but it will put an end to those "I used 1 billion tokens and spent $4" reports.
The problem with DS Flash/Pro is that they are extreme reasoning heavy and step heavy. Step = cache hit. Reasoning = output hit. So the impact on those price increases will be felt much stronger.
I think that Flash is still a usable model but Pro is DOA... Even before the price difference between Flash and Pro, vs the intelligence / problem solving / tool calling did not make sense. But now that gap has widen even more. And there are just too many competitors models now close to that Pro price range.
Especially when we compare that competitive models offer subscription services that easily cut down the token price by 1:10. That makes Pro especially a bad value.
We shall see what the 3th party market is going to do, but i suspect that prices will be increased. If the argument was that DeepSeek increases price as they lack capacity, a company with access to billions, other 3th party providers that need to rent and have less optimized infrastructures will increase prices. Especially if they get hit hard with people moving around.
Its like we always see the same issue with popular models.
* GLM 5.2 is good, capacity issues, API price up, subscription heavy nerfs. * Kimi K3 is good, capacity issues, API price up, subscription heavy nerfs. * DeepSeek V4 GA is good, capacity issues, API price up * OpenAI GLM 5m, 10m active users. Subscription usage is sneakily tightened more and more. * Anthropic Opus too popular, ...
That is the main issue. The AI users are people who actively easily move between companies. Pushing peak loads to each unprepared company, releasing load on the "less desired". And round we go ...
Opencode said they are working on matching the old prices using their own inference.
Right now, they give 4100 credits for Luna and 63 000 for Deepseek on their prepaid plan (both are 2x)
How could that possibly work? Deepseek was undercutting every other provider by an order of magnitude on cached tokens.
Do they just set a super low caching time and hope that drops effective cache rates low enough? Do all other providers somehow overcharge by that much? Are they just going to sell it as a loss leader?
> Do all other providers somehow overcharge by that much?
This, I think. Cached inputs have an opportunity cost (keeping the KV cache until use) but a hit is basically free. “Basically” - if the cache is offloaded to system RAM or NVMe there’s some scheduling overhead.
From a consumer viewpoint a more interesting metric than the raw costs is
from a personal standpoint rather than a per-provider one (e.g. if OpenRouter is blindly dispatching your requests you might have a bad time).> if OpenRouter is blindly dispatching your requests
This can somewhat be the case, depending on your config. I updated mine to make DeepSeek high priority because I was having a lot of cache misses and reliability issues with the default (cheapest (at face value)) providers, and cost was actually higher overall than anticipated. Was smooth sailing from then; might have to tweak things again now pricing has changed though.
System RAM and/or NVMe storage still has a real cost. And swapping out the context between VRAM and system RAM / NVMe still consumes bandwidth.
I don't have a clue on what the real cost to inference providers comes out to, but it seems really weird that there would be such a big gap, in what should be a pretty competitive market.
So, pulling some numbers from my setup - DSv4 Flash with an fp8 KV cache:
That's about 10% over max model len, but to make the numbers round we can call 8GB the upper bound used by the cache for a single sequence. The actual average sequence length is going to be lower, but providers might be using a bf16 KV cache instead of an fp8, so 8GB is probably a reasonable placeholder.A x16 PCIe 5 link has a theoretical per-direction per-link bandwidth of ~64GB/s, so if you've got 8 cards running you should have 8*64GB/s = 512GB/s total throughput from the root complex (some topology restrictions apply). The math works out such that we can transfer the KV cache for 64 sequences per second.
But here's the thing: we don't need to transfer the KV cache for each token. The KV cache works for the entire sequence, and we're generating thousands of tokens (thinking lol) for that sequence for each request. So the KV cache transfer overhead gets amortized over those thousands of tokens (which takes several seconds to generate, even at 100 tg/s).
The actual numbers need e.g. concurrency and such to plug in, but I think the bandwidth concerns are fairly negligible.
I doubt they will match old cache read pricing- that’s most important in agentic coding.
CXL might save us. All the world's old DDR4 to the rescue. Either per box, where the job has to route back, or network attached, where there's now a pool of absurdly fast temp storage.
About 3x increase. Luna is now a much better deal. Hope they don't increase their prices in response.
Luna has one vendor and they can change the price any time. Deepseek is open and has dozens of vendors competing to serve it.
Luna vs v4 Flash, sure.
But DeepSeek v4 Pro is a far more capable model and still cheaper than anything that it competes with, from what I can see.
Why not? They probably only offered those deals because of DeepSeek aggressive pricing. Now that DeepSeek is 3x more expensive it's time to revert those discounts.
Tibo said it’s permanent, whatever that means
Probably means it's not intended as temporary, but not that the price will never change again
It doesn't make any sense unless they are going to exit from inference market. They will be literally one of the costliest option (by output, for flash) if use openrouter as source.
Makes sense, basically increased price for peak hours when they don't have enough infra to serve everyone. Can expect the return of prices back.
Ah well. I spend about $5/month with Deepseek, so now I’ll have to find room in my budget for $15. Might have to tip my barista less or something.
non-x link: https://api-docs.deepseek.com/quick_start/pricing/ (https://web.archive.org/web/20260813114752/https://api-docs....)
This now places deepseek flash v4 from DeepSeek themselves at higher prices than openrouter (depending on caching). Will be interesting to see if third party prices remain the same.
It does seem to me that DeepSeek themselves aren't so much interested in being a service provider. They do it, and offer the service, but their pronouncements seem to be that they're more interested in being for now closer to a research lab with a longer term play for something more dramatic later.
the neat thing about the models being open-weights is there's a dozen other providers on OpenRouter still at the old price, or lower
Unfortunately in this case that's not true at all, there was no provider with genuinely close or the same effective prices (mostly based on cache hit cost) to old v4 flash or v4 pro. People have this misconception that other providers must be much cheaper than the official one in case of open weight models.
If you check on OpenRouter, some other providers serve V4 Flash at seemingly cheaper normal input/output tokens rates, but with a huge caveat: they have at least a 5x increase of the cache hit cost of the official API, some have a 10x+. No provider comes close to Deepseek's old low cache prices, and cache is 90%+ of what matters in agentic sessions.
Closest comparison:
- Deepseek: $0.14/$0.28 with $0.0028 cache hit cost for official API
- DeepInfra: $0.08/$0.18 (cheaper base rates!) with $0.016 cache hit (almost 6x!! Deepseek's current cache cost)
Another great example is Kimi K3, official API is $3/$15 and the cheapest provider on OpenRouter is $2.8/$14, only a tiny difference.
The question is if they'll increase prices as well.
https://xcancel.com/deepseek_ai/status/2087864589895798968