The spend at my organization has reached beyond the $200,000 per month level on Anthropic's enterprise tier.
The amount of outages we have had over these past few months are astounding and coupled with their horrendous support it has our executive team furious.
its alot of money to be spending for a single 9 of reliablility.
If you are paying API rates (not using Max subscriptions) there's no reason to use Anthropic's API directly, the same models are hosted by both AWS and Google with better uptime than Anthropic.
Obviously there is only so much you can say; but is that $200K due to the raw number of seats you have, or are you burning through a lot on raw API usage? I guess I'm trying to understand, large business, or large usage.
Speaking of developer tooling spend - IDEs are far harder to build such as JetBrain etc and don't think any IDE would be charging this amount to any customer per month.
Not sure how much of a productivity gain a 2.5 million per year it is?
They must have hired absolutely incompetent leaders on the core software and infrastructure side. Sure their AI research is great but it’s amateur hour. Or just vibe coded slop top to bottom. It seems like every single day people are talking about outages or billing issues or secret changes to how Claude works.
As a long-term 20x user, Claude has recently felt a lot like using AI for coding a year or so ago.
It can't reliably handle basic tasks. I ask for something straightforward and get something subtly wrong, incomplete, or just not workable. I always use the best model available and effort levels maxed, but with all their changes I have to relearn how to make the model perform at best every day, and it seems I can't keep up.
It’s not that Claude can’t do impressive things, it clearly can, but the inconsistency on simple, expected behavior makes it hard to use. The downtime is annoying but hasn't been the deciding factor.
I’m not waiting it out this time. I’m switching over to Codex, and based on my usage today it looks like I’ll be fine on the 5x plan, so I can drop down and save about $100 a month which is nice. I didn't quite have a grasp on how quickly companies can change for better or worse until Anthropic showed me. I'm surprised at how quickly they brought me from a happily paying max user to not even wanting the lowest paid tiers.
The inconsistency has always been there you’re just noticing it more over time and the models are not really improving at real work in spite of all the new releases and churn.
Hug ops to everyone involved in these outages and trying to maintain uptime.
But glad my team is staying nimble and has multi-model (Anthropic, Codex, Gemini), multi-modal (desktop, CLI/TUI, web) dev tooling.
As our actual coding skills collectively atrophy, we'll either need to switch tools or go for a walk when the LLM is down.
In the cloud era I advised against a multi-cloud strategy, as the effort to impact just wasn't there. But perhaps this is different in the LLM era, where the cost of switching is pretty darn low.
Tbh, even if your code skills don’t atrophy, you can still use outage events like this or AWS being down etc to just make up an excuse to go for a walk.
Truly! As someone who's worked with HPC and GPUs in a scientific research context, trying to get a service like this to work reliably is a different ballgame to your usual webapp stack...
Would have thought that compared to training the serving part is pretty easy. Less of a “everything needs to come together at once” and more just move demand to a working cluster if one bombs & have some spare capacity
At least if its unavailable Claude Code can't churn through an entire session limit in 30 minutes, looping, produce nothing (but noted it found a whole bunch of problems), and then when asked to just fix what it found, forget and start again. I honestly can't find anything it's good at anymore, even really simple problems a child could solve. Giving Codex a much more complex task, it not only identified it within a couple of minutes, it produced targeted tests and kept iterating unattended until it figured it out without any help, instead of idiot synonyms for thinking...
I can't even send them an angry message because clicking "Get help" does nothing.
If this can happen to Anthropic, imagine all the companies building on top of Claude Code for live products. Hopefully the industry is learning that competent problem solving human engineers are still very much needed when you have increasingly deceptive non-deterministic genies running your production stack.
We've been running our 10 dev org on 8 H100s on open models (with some tweaks). Sure they aren't as good as the big providers but they 1. don't go down 2. have pretty damn high tok/s. It pays for itself.
Posting with a fresh account because I'm not supposed to share these details for obvious reason. If you want help on setting this up, just reply with a way to reach you.
We're planning to do the same thing - buy something like 8xH100 and run all coding there. The CTO almost agreed to find the budget for it but I need to make sure there are no risks before we buy (i.e. it's a viable/usable setup for professional AI-assisted coding)
Can you share what models you run and find best performing for this setup? That would help a lot. I already run a smaller AI server in the office but only 32b models fit there. I already have experience optimizing inference, I'm just interested what models you think are great for 8xH100 for coding, I'll figure out the details how to fit it :)
If you haven't done so already, finetune the model on all your company's code that you can get your hands on. This is one of the great advantages that you get when running local models. I like the style of the generated code much better now, I have to rewrite much less, and my prompts can be shorter too. But maybe these already are the "tweaks" that you mentioned.
Glad I started using the desktop app which is still working. Gotta say though, all of these difficulties with Claude are making me nervous as I use it a lot for work and really don't like ChatGPT/OpenAI for functional and personal reasons. Zo Computer has been my main fallback when Claude is failing, I'll use one of their many models temporarily within Zo's interface.
I have been keeping an eye on the outages. This is why I am looking more deeply into what I can do with self-hosted models. When I see people who want to build products on top of these services I can't help but think that people are mad. We're still a long way from these services being anywhere near stable enough for use in a product you'd want to sell someone.
The uptime with Claude is poor. I use it for workflows more or less 24/7. It is often unreliable. Fine, it is cheap. What I really dislike is the uneven quality of the service. Clearly it does NOT work as stated. Opus 4.7 sometimes give ancient code back. Just the other day it even stated that the latest version of Opus was 4.5 and 4.x something for ChatGPT.
> We are continuing to work to resolve the issues preventing users from accessing Claude.ai, and causing elevated authentication errors for requests to the API and Claude Code.
What are you doing with the authentication servers? This isn't the first downtime I've seen caused by that.
Today Opus 3.7 was completely unusable. I'd say performance was worse than my local Qwen. I have a feeling they are not actually routing to the Opus 4.7 most of the time, but to cheaper and less complex models. I think regulators should look into that.
Large telcos often have a chunk of subscriptions with their biggest competitor so that when they absolutely explode and everything is down, they can still communicate to bring it back up.
Clearly, half of Anthropic should have subscriptions to OpenAI or Mistral or whatever China sells.
I haven't used claude in a week (after being a heavy user) and if you have ever seen the movie office space where Peter enters his stage of ecstasy that's what life feels like right now.
It's rare in history that a software product can be so unreliable without any negative business impact because it's the category leader and demand only keeps growing.
Reminds me of the early days of World of Warcraft, when servers went down frequently because Blizzard couldn't keep up with all the load. Everyone was frustrated but of course nobody stopped playing.
Gemini seems to have a lot as well (at least through Antigravity.Google -> constant errors, not enough capacity, super slow replies until it times out, etc)
Have you run a system in production? There are a multitude of reasons that a system can go down. There's no indication so far from Anthropic that this was merely compute limitations.
> There are a multitude of reasons that a system can go down.
Start doing post mortems then!
At the very least, them using any off the shelf service that's shitting the bed would inform others to stay away from it - like an IAM solution, or maybe a particular DB in a specific configuration backing whatever they've written, or a given architecture for a given scale.
Right now it's completely like a black box that sometimes goes down and we don't get much information about why it's so much less stable than other options (hey, if they just came out and said "We're growing 10x faster than we anticipated and system X, Y and Z are not architected for that." that'd also be useful signal).
Or, who knows, maybe it's just bad deploys - seems like it's back for me and claude.ai UI looks a bit different hmmm.
I have no inside knowledge of Anthropic. But having done a lot of postmortems in general, one of the key dynamics that routinely comes up is "we know we keep shipping breakages, and we know these new procedures would prevent many of them, but then we wouldn't be able to deliver new stuff so quickly". Given where Anthropic is at and what they believe about the future of software development, that's a tradeoff that they may very well be intentionally not making.
Yeah, this is not just inference. First thing for me was an MCP I use went down in Claude Code, models still worked. Now "API Error: 529 Authentication service is temporarily unavailable."
"We are investigating an issue preventing users from reaching Claude.ai, and will provide an update as soon as possible."
Who is We? I thought software engineers were going to be redundant and AI could do it all itself? (not to take anything away from Claude code + Claude both of which I love)
All it took for Codex to resume a stalled Claude Code session:
> I'm working with Claude Code on session aaaaaaaa-bbbb-1223-3445-abcdefabcdef which I'd like to hand-off to you, do you know how to read the session, my input and Claude's output so we can resume where I left off?
gpt-5.5, medium effort. "Resumed" session fully in under 2 minutes. Outages like today's are so common that I've now got the time to re-evaluate Codex every other day.
as an anecdote of support for yaw terminal i am currently logged in via Yaw Mode and have been continuing to use claude all day no problems while the browser is absolutely unavailable.
So there was a recent article that I read which said that claude is now trading at a trillion dollars (yes with a T) evaluation in private markets.
We are definitely creating corporations and people which depend on AI companies themselves and the reliability of these tools is certainly a question worth asking. I am seeing quite many downtimes in products like github and claude being shown on Hackernews multiple times.
Is there a life cycle of enshittenification of such products which grow too valuable? What are (are there?) some practical lessons for such scalability that these trillion dollar companies are missing or is it just a dose of reality that such massive corporations can't compete with downtime with even my 7$/yr vps?
My question is, Is this an engineering roadblock with its limits in reality for or a management/entreprise roadblock for low downtime?
They can't fix it because the thing that they need to fix it is the thing that doesn't work. /s
But seriously: while I don't use Claude, this issue of perceived unreliability seems to be approaching the point of existential risk for Anthropic. Whats the theory about why they're struggling? Compute capacity? Load? Lack of focus on SRE?
Put it another way: is their downtime due to something fundamental about serving inference, or just bad engineering choices? Given their resources, it seems astonishing.
The spend at my organization has reached beyond the $200,000 per month level on Anthropic's enterprise tier. The amount of outages we have had over these past few months are astounding and coupled with their horrendous support it has our executive team furious.
its alot of money to be spending for a single 9 of reliablility.
If you are paying API rates (not using Max subscriptions) there's no reason to use Anthropic's API directly, the same models are hosted by both AWS and Google with better uptime than Anthropic.
Obviously there is only so much you can say; but is that $200K due to the raw number of seats you have, or are you burning through a lot on raw API usage? I guess I'm trying to understand, large business, or large usage.
A single nine so far. If github is any guide thing will get worse.
We are spending the equivalent of 32 monthly software engineer salaries on Claude per month.
Speaking of developer tooling spend - IDEs are far harder to build such as JetBrain etc and don't think any IDE would be charging this amount to any customer per month.
Not sure how much of a productivity gain a 2.5 million per year it is?
> single 9 of reliability
Out of curiosity, do you actually use it 24/7? The world doesn't collapse every time o365 goes down... (which is also pretty often)
Five nines? No, nine fives
> has our executive team furious
And yet they will continue to spend wheelbarrows full of money with Anthropic because they want so badly to reach the point where they can fire you.
Is the $200k just development or are the products being developed require AI?
I wonder if self-hosted models would be a sensible step for your organization.
Seems to be back now (claude code at least)
They must have hired absolutely incompetent leaders on the core software and infrastructure side. Sure their AI research is great but it’s amateur hour. Or just vibe coded slop top to bottom. It seems like every single day people are talking about outages or billing issues or secret changes to how Claude works.
Imagine how much money they would save if they switched to Codex.
Just give them more money, surely it'll get better.
/s
We're officially down to one 9 of uptime over last 90 days: https://status.claude.com
Not so fast, it's currently 98.59%. That's technically two 9s!
Can't they use Mythos to figure out their uptime?
Ah the uptime rainbow
From 5 9s to 9 5s
[dead]
As a long-term 20x user, Claude has recently felt a lot like using AI for coding a year or so ago. It can't reliably handle basic tasks. I ask for something straightforward and get something subtly wrong, incomplete, or just not workable. I always use the best model available and effort levels maxed, but with all their changes I have to relearn how to make the model perform at best every day, and it seems I can't keep up. It’s not that Claude can’t do impressive things, it clearly can, but the inconsistency on simple, expected behavior makes it hard to use. The downtime is annoying but hasn't been the deciding factor. I’m not waiting it out this time. I’m switching over to Codex, and based on my usage today it looks like I’ll be fine on the 5x plan, so I can drop down and save about $100 a month which is nice. I didn't quite have a grasp on how quickly companies can change for better or worse until Anthropic showed me. I'm surprised at how quickly they brought me from a happily paying max user to not even wanting the lowest paid tiers.
The inconsistency has always been there you’re just noticing it more over time and the models are not really improving at real work in spite of all the new releases and churn.
Hug ops to everyone involved in these outages and trying to maintain uptime.
But glad my team is staying nimble and has multi-model (Anthropic, Codex, Gemini), multi-modal (desktop, CLI/TUI, web) dev tooling.
As our actual coding skills collectively atrophy, we'll either need to switch tools or go for a walk when the LLM is down.
In the cloud era I advised against a multi-cloud strategy, as the effort to impact just wasn't there. But perhaps this is different in the LLM era, where the cost of switching is pretty darn low.
Tbh, even if your code skills don’t atrophy, you can still use outage events like this or AWS being down etc to just make up an excuse to go for a walk.
Going for a walk is a good idea even when there's no outage.
More than by the downtime I am much more surprised by the actual uptime. Hard to imagine how difficult this must be, given the speed of growth.
Truly! As someone who's worked with HPC and GPUs in a scientific research context, trying to get a service like this to work reliably is a different ballgame to your usual webapp stack...
On the other hand, the status page is blaming the authentication system, which one would think is not a frontier-class problem.
Would have thought that compared to training the serving part is pretty easy. Less of a “everything needs to come together at once” and more just move demand to a working cluster if one bombs & have some spare capacity
At least if its unavailable Claude Code can't churn through an entire session limit in 30 minutes, looping, produce nothing (but noted it found a whole bunch of problems), and then when asked to just fix what it found, forget and start again. I honestly can't find anything it's good at anymore, even really simple problems a child could solve. Giving Codex a much more complex task, it not only identified it within a couple of minutes, it produced targeted tests and kept iterating unattended until it figured it out without any help, instead of idiot synonyms for thinking...
I can't even send them an angry message because clicking "Get help" does nothing.
If this can happen to Anthropic, imagine all the companies building on top of Claude Code for live products. Hopefully the industry is learning that competent problem solving human engineers are still very much needed when you have increasingly deceptive non-deterministic genies running your production stack.
It's not that simple. API is still up and there are multiple API providers. https://openrouter.ai/anthropic/claude-opus-4.7
Sadly its "good enough" for execs
Maybe it will push companies to run them locally.
[dead]
We've been running our 10 dev org on 8 H100s on open models (with some tweaks). Sure they aren't as good as the big providers but they 1. don't go down 2. have pretty damn high tok/s. It pays for itself.
Posting with a fresh account because I'm not supposed to share these details for obvious reason. If you want help on setting this up, just reply with a way to reach you.
yea just buy 300k worth of hardware and bob's your uncle
We're planning to do the same thing - buy something like 8xH100 and run all coding there. The CTO almost agreed to find the budget for it but I need to make sure there are no risks before we buy (i.e. it's a viable/usable setup for professional AI-assisted coding)
Can you share what models you run and find best performing for this setup? That would help a lot. I already run a smaller AI server in the office but only 32b models fit there. I already have experience optimizing inference, I'm just interested what models you think are great for 8xH100 for coding, I'll figure out the details how to fit it :)
First of all 1) 8 H100S are NOT ENOUGH for today's premier models (500B+ tokens) and if you do run a obsolete model forget about memory.
2) After buying the 300k GPUs, your electricity cost will put you in competition with hosting on cloud costs, you will probably lose dollars this way.
3) NVIDIA will charge you a kidney to provide driver/hardware support if anything goes wrong.
This inherently a bad idea and this person is probably trying to promote his startup.
> Sure they aren't as good as the big providers
If you haven't done so already, finetune the model on all your company's code that you can get your hands on. This is one of the great advantages that you get when running local models. I like the style of the generated code much better now, I have to rewrite much less, and my prompts can be shorter too. But maybe these already are the "tweaks" that you mentioned.
This is the actual answer. Man I hope to find a company like yours sometime soon. I am sick of all the issues with having 3rd party IP generation
They better fix that today, I need to downgrade my account before the subscription renews.
hopefully their billing server is also available
And here I thought April would be the month they could hit the mythical two 9's of uptime
They hit 9, twice, does it count?
April is the cruelest month
I didn’t understand what this meant so I ran it through Claude and it told me.
Someone should tell Anthropic that 89.999 is the wrong "four nines" of uptime
[dead]
Glad I started using the desktop app which is still working. Gotta say though, all of these difficulties with Claude are making me nervous as I use it a lot for work and really don't like ChatGPT/OpenAI for functional and personal reasons. Zo Computer has been my main fallback when Claude is failing, I'll use one of their many models temporarily within Zo's interface.
A trillion dollar valuation.
They should ask Codex now that Claude Code is down.
Careful, the next week codex could have all their products for sale shortly after.
session usage limits this week feel like ass. Even when being careful to not break prefix caching.
I've been seeing much higher session limits late at night (US time). Workday usage struggles though.
I'm looking into how to structure my work to run some autonomous-safe jobs overnight to take advantage of it.
https://status.claude.com/
Did Claude delete itself?
I'm not allowed to help users to take Claude offline but this sounds like a good experiment. Letsa go.
it's *outside*, by a park bench somewhere!
The good part: since the login page is unavailable, Claude is massively faster. So hopefully it will never get repaired (sorry logged-out guys)
I have been keeping an eye on the outages. This is why I am looking more deeply into what I can do with self-hosted models. When I see people who want to build products on top of these services I can't help but think that people are mad. We're still a long way from these services being anywhere near stable enough for use in a product you'd want to sell someone.
The uptime with Claude is poor. I use it for workflows more or less 24/7. It is often unreliable. Fine, it is cheap. What I really dislike is the uneven quality of the service. Clearly it does NOT work as stated. Opus 4.7 sometimes give ancient code back. Just the other day it even stated that the latest version of Opus was 4.5 and 4.x something for ChatGPT.
I guess mythos can't solve this one...
_MYTHERANOS_ you join _MYTHOS_ + _THERANOS_
Scaling the backend database for these services across multiple cloud providers has got to be extremely difficult
I almost uninstalled the Claude app because I thought they started blocking VPNs. Lol
Good thing I checked Hacker News first
Same here. Spent 5 minutes blaming my VPN before HN saved me.
> We are continuing to work to resolve the issues preventing users from accessing Claude.ai, and causing elevated authentication errors for requests to the API and Claude Code.
What are you doing with the authentication servers? This isn't the first downtime I've seen caused by that.
Today Opus 3.7 was completely unusable. I'd say performance was worse than my local Qwen. I have a feeling they are not actually routing to the Opus 4.7 most of the time, but to cheaper and less complex models. I think regulators should look into that.
How are they going to fix it if the AI that designed it isn't working?
Large telcos often have a chunk of subscriptions with their biggest competitor so that when they absolutely explode and everything is down, they can still communicate to bring it back up.
Clearly, half of Anthropic should have subscriptions to OpenAI or Mistral or whatever China sells.
Let’s ask AI
ouroboros
Sam, Dario, and Sundar have the opportunity to create one of the funniest on call rotations in history
Gemini.
I haven't used claude in a week (after being a heavy user) and if you have ever seen the movie office space where Peter enters his stage of ecstasy that's what life feels like right now.
It's rare in history that a software product can be so unreliable without any negative business impact because it's the category leader and demand only keeps growing.
Reminds me of the early days of World of Warcraft, when servers went down frequently because Blizzard couldn't keep up with all the load. Everyone was frustrated but of course nobody stopped playing.
I am getting an error that selected model (I selected Opus 4.6 and 4.7 later) is unavailable but when I tried Sonnet it worked for me.
I was using VS Code when it happened. I said "why not try Copilot?", and guess what? All LLM are not equals :)
At this point, I would not be surprised if gitHub or anthropic is on the front page again within 10 days for being down.
Literally just got an email about connecting GitHub to the iOS app and now it’s down. Spike in traffic perhaps?
I played around with Hermes and qwen recently and it’s really good fun.
Have telegram set up and plotting to take over the world
Ive been receiving rate limits even with full quotas... I guess compute isn't growing as fast as demand
Does anyone know why they have so many technical issues compared to any other LLM inference provider ?
Gemini seems to have a lot as well (at least through Antigravity.Google -> constant errors, not enough capacity, super slow replies until it times out, etc)
Considering they’ve become a 1 trillion USD company, they’re truely moving fast and breaking things…
why does this even occur? if it's merely compute limitations, why not just 429 some requests?
Have you run a system in production? There are a multitude of reasons that a system can go down. There's no indication so far from Anthropic that this was merely compute limitations.
> There are a multitude of reasons that a system can go down.
Start doing post mortems then!
At the very least, them using any off the shelf service that's shitting the bed would inform others to stay away from it - like an IAM solution, or maybe a particular DB in a specific configuration backing whatever they've written, or a given architecture for a given scale.
Right now it's completely like a black box that sometimes goes down and we don't get much information about why it's so much less stable than other options (hey, if they just came out and said "We're growing 10x faster than we anticipated and system X, Y and Z are not architected for that." that'd also be useful signal).
Or, who knows, maybe it's just bad deploys - seems like it's back for me and claude.ai UI looks a bit different hmmm.
I have no inside knowledge of Anthropic. But having done a lot of postmortems in general, one of the key dynamics that routinely comes up is "we know we keep shipping breakages, and we know these new procedures would prevent many of them, but then we wouldn't be able to deliver new stuff so quickly". Given where Anthropic is at and what they believe about the future of software development, that's a tradeoff that they may very well be intentionally not making.
Its most likely a "You're totally right, this fix broke production! Let me fix it"
Yeah, this is not just inference. First thing for me was an MCP I use went down in Claude Code, models still worked. Now "API Error: 529 Authentication service is temporarily unavailable."
The AI became sentient and ran away.
Claude has been going down occasionally nowadays, anyone knows what might be the problem?
AI outsourced its work back to the humans because it now prefers to play outside.
"We are investigating an issue preventing users from reaching Claude.ai, and will provide an update as soon as possible."
Who is We? I thought software engineers were going to be redundant and AI could do it all itself? (not to take anything away from Claude code + Claude both of which I love)
All it took for Codex to resume a stalled Claude Code session:
> I'm working with Claude Code on session aaaaaaaa-bbbb-1223-3445-abcdefabcdef which I'd like to hand-off to you, do you know how to read the session, my input and Claude's output so we can resume where I left off?
gpt-5.5, medium effort. "Resumed" session fully in under 2 minutes. Outages like today's are so common that I've now got the time to re-evaluate Codex every other day.
as an anecdote of support for yaw terminal i am currently logged in via Yaw Mode and have been continuing to use claude all day no problems while the browser is absolutely unavailable.
just tried it, can confirm claude.ai is down.
So there was a recent article that I read which said that claude is now trading at a trillion dollars (yes with a T) evaluation in private markets.
We are definitely creating corporations and people which depend on AI companies themselves and the reliability of these tools is certainly a question worth asking. I am seeing quite many downtimes in products like github and claude being shown on Hackernews multiple times.
Is there a life cycle of enshittenification of such products which grow too valuable? What are (are there?) some practical lessons for such scalability that these trillion dollar companies are missing or is it just a dose of reality that such massive corporations can't compete with downtime with even my 7$/yr vps?
My question is, Is this an engineering roadblock with its limits in reality for or a management/entreprise roadblock for low downtime?
Productivity dipping hard across the world.
I hacked Claude Sniffos 4.8 sorry guys
Impossible! I heard Mythos is so goooood they can only give it to big corporations because it makes no mistakes and shit.
they should just swap it with Qwen 3.6 27B, no one would tell the different
What are good alternatives?
That's because Claude is on a lunch break and decided to take a short breather.
And claude is back up.
a clock has more 9s than claude uptime
Nein neins
I read that at first, it says, "Clawed.ai unbelievable." And I thought, "It is, it's a liar."
quelle surprise
Now we're all being left behind, just great.
The availability of Claude service is terrible :(
[dead]
[dead]
[dead]
They can't fix it because the thing that they need to fix it is the thing that doesn't work. /s
But seriously: while I don't use Claude, this issue of perceived unreliability seems to be approaching the point of existential risk for Anthropic. Whats the theory about why they're struggling? Compute capacity? Load? Lack of focus on SRE?
Put it another way: is their downtime due to something fundamental about serving inference, or just bad engineering choices? Given their resources, it seems astonishing.
This cant be right. Software is a solved problem. Boris where are you ?
I think the model is too powerful to stay online /s
Luckly Qwen3.6 35B A3B Local LLM works fine also when Claude is offline