GitHub Community
GitHub Copilot Pro+ "rate limited" for hours (user_weekly_rate_limited) #192485
Replies: 71 comments 101 replies
-
|
💬 Your Product Feedback Has Been Submitted 🎉 Thank you for taking the time to share your insights with us! Your feedback is invaluable as we build a better GitHub experience for all our users. Here's what you can expect moving forward ⏩
Where to look to see what's shipping 👀
What you can do in the meantime 💻
As a member of the GitHub community, your participation is essential. While we can't promise that every suggestion will be implemented, we want to emphasize that your feedback is instrumental in guiding our decisions and priorities. Thank you once again for your contribution to making GitHub even better! We're grateful for your ongoing support and collaboration in shaping the future of our platform. ⭐ |
Beta Was this translation helpful? Give feedback.
-
|
The copilot service is ridiculous to say the least and now i have been rate limited for requesting I think 20 questions in a morning. |
Beta Was this translation helpful? Give feedback.
-
|
The Same issue here I hit some 5 Hour limit also, now it broke my workflow and the ideas that i have in mind :(. Verry unhapy and I`m on Pro+..... May be I will ditch the Visual Studio and go for the claude code instead. |
Beta Was this translation helpful? Give feedback.
-
|
I get blocked for 80 (!!) hour. Its Thursday. The worst thing is that I cannot even pay to get some additional request. GitHub could make a lot of money by just presenting the payment link in copilot panel. Its hard to understand why they act so stupid. Its a very poor service (only AI answers). I complained several times but did not get any reply. I will switch now to Claude code. |
Beta Was this translation helpful? Give feedback.
-
|
Same rate limited for around 100 hours here... |
Beta Was this translation helpful? Give feedback.
-
|
Por un momento crei que yo me habia equivocado en algo, pero veo que es una queja masiva y una limitacion ridicula de Copilot, yo pago mi suscripcion y ademas pago los excedentes, ni siquiera avisan que se va a acortar. |
Beta Was this translation helpful? Give feedback.
-
|
Here one: openai/codex#17738
Copilot Pro+ rate limit for 2.7 week! GitHub/Microsoft must explain this urgently |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
Also hit the weekly rate limit after 1 days use. But also now no other model is working in VS, just sits there with the three dots doing nothing. |
Beta Was this translation helpful? Give feedback.
-
|
Got this 5 mins after raising a ticket: Thanks for writing in to GitHub Support! I completely understand your frustration with hitting rate limits. The good news is we're actively working to improve this experience for you and all GitHub Copilot users. Your rate limit should reset automatically, but I wanted to provide some context on why this happens and how you can work around it. While Copilot Pro, Pro+, Business, and Enterprise plans have included premium requests, there are still global rate limits that apply to all Copilot plans which are unrelated to your premium request balance or budgets. These protect the stability of the service while capacity is limited and cannot be modified by GitHub Support. Our biggest challenge right now is securing more capacity for our premium models. Some of these models aren't running on our primary infrastructure, and we're working diligently with cloud providers to increase availability. The Copilot rate limits will often be triggered by a significant amount of tokens being used at once. Here are a few ways this can happen:
A simple way to reduce the amount of tokens being used would be to regularly start a new conversation when moving onto a new topic. As a conversation with GitHub Copilot grows, all the prior context and messages are still included which can result in many tokens being used. Rapidly iterating and complex AI features, like agentic AI can use a lot of tokens rapidly. Using premium models with agentic features can quickly exhaust rate limits. You might need to change models depending on the specific task you're working on, for example using the default model for most tasks and then starting a new conversation targeted on specific complex tasks with a more advanced model. See [Choosing the right AI model for your task] in the GitHub Docs for more details. I hope these pointers are helpful. Our team is eager to roll out greater rate limits, and is constantly working on increasing capacity for all GitHub Copilot users. While third-party provisioned models have more limited resources, we're actively working to improve this situation too. We appreciate your patience as we Accelerate Performance across our AI-powered developer platform. |
Beta Was this translation helpful? Give feedback.
-
|
if its right sorry microshit |
Beta Was this translation helpful? Give feedback.
-
3 easy steps to solve the rate-limit problem:1:
2:
3:
I blocked GitHub from my credit card, I will not pay a single cent, I'm done with Fyoumicrsoft. |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
-
|
да это жопа полная раньше просто рейт лимит был сейчас недельный лимит хоть по 350$ в месяц туда сливаю - надо что то менять и искать альтернативы |
Beta Was this translation helpful? Give feedback.
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
-
|
If this is will not be fixed, the better way - move to OpenRoute |
Beta Was this translation helpful? Give feedback.
-
|
Microsoft demonstrating how to speedrun losing your entire user base in a week. |
Beta Was this translation helpful? Give feedback.
-
|
https://github.com/orgs/community/discussions/192707#discussioncomment-16588856
|
Beta Was this translation helpful? Give feedback.
-
|
deleted image |
Beta Was this translation helpful? Give feedback.
-
|
What was a tool, has become "owned" by Github and they decide if you can use it or not.... Even though we have paid for it and even through we are willing to pay for premium requests once the quota is used up for the month, they have made a tool something you feel like is not a tool anymore but something you have to wait on for it to turn up... like a taxi driver you are waiting on when you are in a hurry to the airport and does not turn up on time and instead takes its time and is late and then takes 15 pit stops before its destination >> and expects full fare to be paid!!! How are we supposed to use this for business when you just turned it off! I am sure your (Github) terms and conditions were written to allow you to do this, but you are essentially killing your client base who are serious about using and paying for it...... Is this a clear "we dont care about you - just f%^&* off and find somewhere else" signal.... You can not get around the weekly rate limit, not even pay for it to keep going >>>> YET ANYWHERE ELSE THERE IS NO SUCH THING that would block you constantly due to a weekly rate limit > you pay for it..... WHY WOULD THEY KILL THEIR CUSTOMERS FAITH AND TURN THEMSELVES INTO THE UNRELIABLE TOOL THAT SAYS WAIT x MINUTES AND THEN AFTER 30 SECONDS IS RATE LIMITED AGAIN!! SERIOUSLY! I CAN ONLY IMAGINE THE AMOUNT OF PEOPLE EXITING COPILOT NOW AS YOU HAVE FORCED ME TO DO THAT IN ORDER TO KEEP MY BUSINESS WORKING................ MAYBE I SHOULD FIND IN YOUR TERMS WHERE I CAN CLAIM DAMAGES FOR NOT BEING ABLE TO DO WORK AND CAUSING ISSUES AS A RESULT OF WHAT I HAD ALREADY PAID FOR.... NO WAY OUT AND STUCK AND FORCED TO LITERALLY HATE THIS COMPANY AND FIND ANOTHER SOLUTION.... SQUEEZED OUT?? WHY WOULD THEY DO THAT? if you need to charge more then why are you not doing that instead of squeezing customers to go elsewhere.... obviously not interested in customers and have another agenda, that is the only thing that makes sense.... When the tool does work, it is fast and works timely, so there is no degradation of the product that i had noticed previously at all so why have they done this??? |
Beta Was this translation helpful? Give feedback.
-
|
when i checked what you said me eyes burned they are so many text but its true |
Beta Was this translation helpful? Give feedback.
-
|
I guess Microslop is at it again. I'm locked out. |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
This comment was marked as off-topic.
This comment was marked as off-topic.
-
|
Same issue here. Account: hohlfelduglum518-prog, Pro+, hit with Could a GitHub staff member please review and reset? Thanks. |
Beta Was this translation helpful? Give feedback.
-
only for pro+ who do like more credit bcs its a issue with their server
|
Beta Was this translation helpful? Give feedback.
-
|
To whoever is supposedly managing this "premium" service, Let’s get one thing straight: I am paying you $300 to $400 every single month. I pay this exorbitant amount specifically so I don't have to deal with amateur-hour interruptions. Yet, you somehow have the sheer audacity to hit me with a 63-hour and 40-minute rate limit. Are you fundamentally incapable of understanding how a paid subscription works? The fact that a top-tier paying customer is subjected to a 3-day lockout shows a level of technical incompetence in your system architecture that is, frankly, embarrassing. I am paying you premium cash to actually use the service, not to be treated worse than a free-tier user. And your system's automated suggestion to "Consider switching to Auto"? You can keep that completely tone-deaf, useless advice to yourselves. I don't pay hundreds of dollars a month to be mismanaged by a broken algorithm. This isn't a service; it is a complete joke. You are charging premium prices for an utterly garbage user experience. Fix this catastrophic failure immediately, or I will take my money to a competitor who actually knows how to run a functional platform. |
Beta Was this translation helpful? Give feedback.
-
|
Which competitor should you migrate to? Tell us options please |
Beta Was this translation helpful? Give feedback.
-
|
I’ve switched to RunPod Serverless. I’m hitting 200-300 TPS (tokens per second) with the Qwen 3.6 35B-A3B model, and I’ve easily enabled unrestricted internet access. This provides a perfect foundation for my workflow; frankly, I’m able to develop much more efficiently using 'Context OS'—a system where I handle everything from A to Z and build all the prompts myself. I strongly suggest that anyone currently paying $300–$400 a month for external services should consider switching to a dedicated server! The cost of using a 96GB RTX Pro 6000 on serverless for 8 hours a day is approximately $1.69 * 8 * 30 = $405. (Keep in mind, that’s an absurdly high usage estimate. Between waiting for commands and switching to new chats, you’ll likely only have 3–4 hours of active usage per day. Plus, because it delivers outputs 6–7 times faster than Copilot, you can get significantly more work done in much less time.) Additionally, I’ve integrated models like Opus 4.7 via OpenRouter for architectural planning. By combining these architectural blueprints with Context OS, I’m ensuring that Qwen operates seamlessly and efficiently within the workflow. |
Beta Was this translation helpful? Give feedback.
-
|
I openly recommend a more sensible use case to everyone: renting GPUs by the hour from places like Hetzner or Vast. If my calculations are correct, for $400 a month I can use the Qwen 3.6 (0.8-1.2K tps 35-3b) and Qwen 3.5 (0.4K tps 120-12b) models indefinitely for 8 hours a day. I can turn them off at night and avoid wasting money. |
Beta Was this translation helpful? Give feedback.
-
|
You can also use serverless solutions like Openroute, Together.ai and many others |
Beta Was this translation helpful? Give feedback.
-
|
Openrouters or any company's API-based solutions consume an excessive amount of money :D But as you said, if there's a more logical and reasonable solution based on serverless principles, I could use it. Opensource models have become truly intelligent now; there's no need to waste money on these scammers. |
Beta Was this translation helpful? Give feedback.
-
|
I tried using open-source models, but they didn't support tools and it wanted me to hand feed it everything. I'm interested in trying it again. Looks like I might have to. |
Beta Was this translation helpful? Give feedback.
-
|
The Qwen 3.5 and 3.6 series are excellent at tool calling; I think you should give them another shot. Of course, you’ll need at least the 9B model or higher to get the performance you're looking for. |
Beta Was this translation helpful? Give feedback.
-
|
Github copilot pro and its rate limiting after just 2-3 prompts both weekly and daily waiting hours its now literally useless |
Beta Was this translation helpful? Give feedback.
-
|
Github Pro user here. Ran into my weelky session after 15 or so prompt using a 0x model (GPT-4.1). This is absolutly unacceptable. Microsoft just destroyed the only "Copilot" product people actually liked. The fact this issue happened so sudently with no warning or notice is unacceptable, the capacity and/or money issue didn't suddenly appear, it was slowly building and Github & Microsoft didn't have the forsight to do anything about it. They could have done: warned users, create other plans, implement less aggressive limilts gradually, limited larger models... just name a few. |
Beta Was this translation helpful? Give feedback.
-
|
I found the perfect solution (for me). Not advocating for it, but just to let you know. |
Beta Was this translation helpful? Give feedback.
-
|
What kind of nonsense is this, man! You're even putting a "rate limit" when I want to use the openrouter model, what a disgrace! Fuck you Microsoft! |
Beta Was this translation helpful? Give feedback.
-
What kind of jokes
|
Beta Was this translation helpful? Give feedback.
-
|
Did Auto work for anyone in this way? |
Beta Was this translation helpful? Give feedback.
-
|
Why do I even pay for a fixed ammount of requests when I get rate limited non the less? |
Beta Was this translation helpful? Give feedback.
-
|
Yeah quite weird, I would get it if you were to get rate limited for certain models, but getting rate limited for all models except auto? Is just weird. Like I went gpt 5.4 but now I also can't use claude. Edit: Nevermind auto doesn't work either lol |
Beta Was this translation helpful? Give feedback.
-
|
This is abnormal — I got rate-limited for 84 hours. I’m leaving Copilot for good. What code assistant would you guys recommend? |
Beta Was this translation helpful? Give feedback.
-
|
I can't seem to find a good alternative honestly. Jetbrains AI assistent is good but, way too pricey. |
Beta Was this translation helpful? Give feedback.
-
|
I will give cursor a try |
Beta Was this translation helpful? Give feedback.
-
Did you give it a try? I'm looking at it now, but hope it's not limited |
Beta Was this translation helpful? Give feedback.
-
|
*Disclaimer : the post is "discussion type" and does not intend harm in any ways to anyone. It merely provides opinionated feedback. The single best reason for me to use copilot was that it offered different models at better rates. Hold on, i know where this is going. copilot targeted "Request Based Pricing" .. meaning if it thought that the request is not over, it did not use my request. so yes it used a lot of requests for Q&A. but copilot never targeted that. so that was inefficient and was expected. It excelled at development. creating tasks itself and then executing them. so overall with this structure, copilot pricing was unbeatable if used efficiently. however, microsoft soon understood that if the model requires editing more files, then it will use more tokens. so in that case users are not the los*rs, but the company is. So, "Introducing rate limits". in disguise of better stability. now the true users are losrs. no matter how much you are paying, you can't get what you paid for. Why?? because if you do get what you paid for, their hidden margins are gone. that's why they say that use auto mode, so they will give you claude haiku 4.5 or gpt models which we know won't do the tasks. so 10% discount, and will use 25% more requests, and ultimately you will use 15% more requests .. and not to mention you are still being rate limited. But there are genuine reasons right?
But in reality, those are actually the problems that are in the brackets. problem solved. (Rate limits) (a bit of sarcasm ahead*)
Good.. Now users can't use the product, but payment is coming through. Revenue.. Revenue.. ✅✅✅ Now, coming to the real solution based on my opinion, All this comes to only one key usage issue.. people using too many tokens on request based plans. then why don't just make it tokens based? track them, show limits transparently. just like claude code does. I think being transparent results in better user satisfaction then degrading user experience in the shadows. |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
This change is bizarre, not only is imposing these new limits a frustrating re-negotiation of the terms of our subscription, the implementation is super flawed. For one thing, I didn't get any sort of indication that there was a rate limit until I hit it, apparently they didn't bother making that work on a few weeks out of date VS Code install. But also, why would the rate limit window be so big? It's one thing to prevent flooding by limiting requests per minute, and I can see having account-level token limits. Heck, you can even rate limit by artificially slowing responses down a bit. But being able to hit a weekly limit and then just... you can't use the tool at all for a week? That's nonsense. Especially since it's easy to hit that limit in a single day! Just last night I decided to try using the agent to build a full project for the first time, and I'm already at my weekly limit? That doesn't make any sense at all. And how is upgrading to a plan that costs four times as much a solution? I apparently can't just buy credits or something, the only way to life the limit is jumping to a much much higher plan tier? I don't need/want this much AI usage all the time, I needed it for this one project and now I'm stuck. EDIT: Also, now that I updated VS Code to a version that fully supports this new nonsense I'm having all sorts of fun new bugs! Agent terminal sessions are super broken, and the agent keeps just spinning trying to get stuff to run and trying to debug phantom issues that are caused by the software not by the code it's writing, which means it's wasting my now very finite request limit going in circles just because the extension has more bugs than a beehive. |
Beta Was this translation helpful? Give feedback.
-
|
The only solution i see is to cancel the subscription and switch to claude code. And then use complimentry models from nvidia using open router and another instance of claude code to make up for the limits. At this point copilot is just playing waiting game. Let people complain, they will accept it eventually. |
Beta Was this translation helpful? Give feedback.
-
|
I've read that GitHub is most likely planning to change to an API call-based model instead of request-based model soon. From the way they're shifting everything so suddenly and drastically I believe it to be true as well. I also think the subscription model tiers will change. This is probably just the beginning, though that's all just my speculation at this point. This doesn't excuse their utter lack of communication or transparency, though, which is also what really disappoints me the most. :( |
Beta Was this translation helpful? Give feedback.
-
|
Well ... and there we have it, usage-based billing... :V |
Beta Was this translation helpful? Give feedback.
-
I have GitHub Copilot Pro+, and was just rate-limited for 16 hours 52 minutes because of an error code "user_weekly_rate_limited". I'm assuming I would not be able to use this for at least the specified time, plus up to a week since the error code did not specify the range of their week definition. Here is what I am going to do as short-term and long-term plan:
It would be great to see more comments either in this post or Github, Reddit, etc. community to share their experience switching from Copilot to other products, such as Claude Code, that'd help me a lot, and I will do so as well when I find better viable options. |
Beta Was this translation helpful? Give feedback.
-
|
With the new pricing model which is basically API usage of $10 or your amount, its basically not a subscription model anymore. And this sucks even more: https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing#model-multipliers-for-annual-copilot-pro-and-copilot-pro-subscribers So what I used today would be given to me over 2 months lol 1 day usage = 2 months with current changes. Claude code has been kinda reasonable and better at fixing bugs for me and $20 is kinda worth it looking at the new changes here |
Beta Was this translation helpful? Give feedback.
-
|
Just happened to me while refactoring code for my website. I was in the last few steps (debugging an api call) and it says I'm rate limited but won't say when I can try again. I've been waiting over 5 hours now. Sorry, you have been rate-limited. Please wait a moment before trying again. [Learn More] |
Beta Was this translation helpful? Give feedback.
-
|
Just 1 request now rate limit. What the hell is all this. |
Beta Was this translation helpful? Give feedback.
-
|
got this mail today Hi there, You're receiving this because you have a Copilot Pro or Pro+ subscription, and we want to notify you of some important updates. GitHub Copilot isn't the same product it was a year ago. It has evolved from an in-editor assistant into an agentic platform capable of running long, multi-step coding sessions, using the latest models, and iterating across entire repositories. Agentic usage is becoming the default, and it brings significantly higher compute and inference demands. To keep up with the way you use Copilot, we're transitioning to usage-based billing, and we want to give you enough time to prepare. If you're also an admin on a Copilot Business or Enterprise plan, you'll receive a separate email covering what's changing for your organization. Last week, we made temporary updates to Copilot individual plans to improve reliability and performance ahead of the broader move to usage-based billing. We will loosen usage limits once usage-based billing is in effect. What's changing on June 1 To help you prepare for usage-based billing, we're launching a new billing preview page in early May that lets you calculate what April's usage would have been under the new billing model before it takes effect. You can find more details on usage-based billing in our blog post and FAQ. We appreciate your patience as we build a more sustainable Copilot for everyone. The GitHub Team |
Beta Was this translation helpful? Give feedback.
-
|
Oh, so they are making it worse than all other products out there, nice |
Beta Was this translation helpful? Give feedback.
-
|
Fantastic, so basically what's the point of the monthly subscription then? The point should be that I pay $10/mo and I get to use whatever I need to use, some months it might be a lot, some it might be nothing at all, and I don't have to think/worry about it. If it's going to be based on total usage, there should be no monthly fee at all, and it should work like the API model that OpenAI uses. I can already put my OpenAI API key into copilot and use that instead, I don't need Github to offer me the same service. And the point of the subscription model from Github's perspective should be that some users use a lot more tokens and some use basically none and they should set the pricing so that it all averages out. If that isn't what they're going to do, then this is the wrong pricing model. Though not offering that as an option really sucks for users, it's one thing for an enterprise account to have a big pool of tokens to budget when they have dozens of employees. But for personal use this is too much overhead to worry about. I just want to give Github $10/mo and have my account features work. I don't want yet another thing I have to manage. At least code completions and next edit suggestions don't use credits, that's the feature that's actually good. But good luck trying to convince me to care about more agent features if I can't reliably use them! Which brings me to another point: If this is the pricing model, fine. But can we please go back to VS Code being VS Code instead of a way to shovel Copilot at us? It's been months since major UX and stability features were part of a VSC update, everything has just been more MCP and agent stuff - all features that require a subscription to actually use. These paid features should not be taking up the entire VSC dev cycle! |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
yeah, Its so horrible, so basically we can make like 27-33 requests/month, like less than 1 a day avg if you use any good model lol |
Beta Was this translation helpful? Give feedback.
-
|
Guys they literally want us to cancel our plans, that's the whole point. They want to move away from single users to enterprise and use resources over there. That's what makes sense to them. That is why all this crap. Aggressively rate limit users and make them change plans or make them cancel so they can move to enterprise customers. The new change they made, basically makes it useless Are you kidding me??? |
Beta Was this translation helpful? Give feedback.
and others
Uh oh!
There was an error while loading. Please reload this page.
-
🏷️ Discussion Type
Bug
💬 Feature/Topic Area
Copilot in GitHub
Body
It's absurd that I pay for GitHub Copilot Pro+ and getting rate limited due to "user_weekly_rate_limited". Today is Tuesday and I had yesterday "Copilot Premium Request
180 requests".
GitHub shifting the blame to users because their infrastructure is useless and instead of limiting enterprises we paying individuals are forced to use "auto" models (GPT) and "ghosting" paying users with error messages like "Sorry, you have been rate limited... user_weekly_rate_limited".
It's hilarious and I strongly consider canceling my subscription.
githubstatus.com intentionally avoid mentioning this while users flooded social media with the same problem. Some users reported well over 200 hours "lockout" which is well over a week.
GitHub basically trolling us.
Beta Was this translation helpful? Give feedback.
All reactions