Github Copilot throws rate limit exceeded error, even after an active subscription!

[qazi9amaan]

Select Topic Area

Question

Body

I've an active subscription for a year but it keeps throwing this error and the Github copilot fails

Here's the subscription

[fahim9778]

@qazi9amaan
Please contact GitHub support as per the last paragraph of this message.

[qazi9amaan]

Is there any other way to contact to GitHub other than creating a ticket? @fahim9778

[iamcjwilliams]

@qazi9amaan Please refer to this post

[lpm0073]

+1
i recently signed up for a paid plan which, until just now, has worked well from inside of VS Code. But, i'm just sitting down to start code, and i got this response on my first prompts. I'd been away from my machine for at least 10 hours, and so the "rate limit" assertion seems misguided.

Edit: in my case the problem appears temporary. i only got the "rate limit" response for a few minutes.

[kkc-tonywu]

Me too...

[JimNayzium]

This is happening to me today. I pay for the Business Co-Pilot not on the free version and I haven't done much of anything differently today than any other day and it jsut start throttling me for what seems like no apparent reason. Any ideas as to why this happens randomly?

[ThimanKrish]

@JimNayzium Just happened for me. I also have a paid version. Tried restarting vscode, open a new chat but no luck. Seems like an ongoing issue for many. I saw there are lots of recent comments in here microsoft/vscode-copilot-release#732

[ThimanKrish]

@JimNayzium Check now. It's back again. Seems like not enough resources for recent demand.

[JDubb-At-Nelnet]

Same here - corporate group license ( I think) . I've been using copilot edits quite a bit - hit the rate limit in one hour of work this morning, locked out for 9 hours. Using mainly the copilot edits feature.

[christopherdude]

Me, as well. I only use it for Visual Studio, haven't used VS in about a week, and suddenly I've exceeded my usage? They have no problem taking my money, but no way to reach any support to figure this out?

[nn1k4]

I am encountering an issue that is affecting my ability to use your plugin. I keep receiving the error "Rate Limit Exceeded," and this happens every day. This significantly hinders my usage of the plugin, which is unacceptable given that I have a paid subscription.

Please look into this matter as soon as possible and provide a solution.

Thank you!

[Hatherall]

Im now getting the same error all of the time, not even using it heavily either, with a paid Pro annual subscription..

[awolfson2]

same

[toolateralus]

same here

[sheisrachel]

same

[gmcosta100]

Same is happening every other time I share the error code from debug file for reviewing.
Now, it did start to impact my productivity. I can't move forward with the system stopping me. Anyone knows how to resolve it?

[simlei]

also not excited about this. I am asking myself if I somehow let my subscription run out after one (!) :Copilot invocation in the vim plugin gives me the rate limit message. Please do better asap.

[taholmes160]

Having the same issue -- this is NOT cool -- im a paying user

[Ozolinsj87]

im also having issue with Error uploading image Error: rate-limit-exceeded. now for two days it still keeps returning that error, I even took Copilot Pro version but still cant upload image

[PieterPaulussen]

Experiencing the same issue in VS Code...

[medichops1]

I am having the same issue. After a couple of chats it returns a rate limit error. It makes it pretty worthless and not worth paying for. I'd rather pay more and get a decent limit as it's able to work with me one just a couple of things before I need to wait an hour to use it again. Really bad.

[AMBudnik]

Same here. I tried everything and still get the Error uploading image Error: rate-limit-exceeded. It's been 5 days now.

[Kipperlenny]

I get this rate limit error all the time... like every second prompt... I am a paying user - how can I pay more to have a higher rate limit?

[nikushx]

Please like this comment if you think this is utterly unacceptable and pledge to cancel the subscription by 1/1/2025 if it is not addressed.

I've been plagued by this and tried for days and weeks to figure out the issue and there is no clear reason why I'm dumping money into an AI that I can't even use.

I am fed up.

inb4 - we've reached out to support and we're kicked around

[nikushx]

@Kipperlenny @AMBudnik @medichops1 @taholmes160

[AMBudnik]

Today the service was offline and now I am able to use the upload without any errors. Hope it works for you as well.

[kumerian]

It's good to see that this bug is still going strong after 2 years....

[charithharshana]

Correct

[nikushx]

GitHub has failed to address this issue that they are effectively stealing $100 we all paid them for a broken service.

Cancel your Copilot subscription and go download Cursor.

Thank me later.

[gmcosta100]

Agreed on cancelling it. No improvement after such a long time, unfortunately.

[kumerian]

Alas, Cursor isn't even close to being a decent replacement. As long as it doesn't integrate into anything, it'll always be limited.

I use VSCode and ADF. Github Copilot perfectly integrates with both of them. Cursor doesn't.

[nikushx]

FWIW - VS Code extensions are entirely usable in Cursor.

[kumerian]

FWIW - VS Code extensions are entirely usable in Cursor.

That's not very useful. It needs to be the other way around.

[nikushx]

FWIW - VS Code extensions are entirely usable in Cursor.

That's not very useful. It needs to be the other way around.

Not sure I understand. You can install GitHub CoPilot in Cursor if you really want...

[EFlynn7778]

I have now been booted to the very back of the line with 11 hour waits.

[newideas99]

Ya can we get more clarification on this

[Jigglefrizz]

I'm also getting rate limited. I've made my own ticket... no reply.

[AwwCoconuts]

After opening a ticket- i received the following reply:
"The Copilot rate limits are typically triggered by a significant amount of tokens being used at once, there are a few ways this can happen:
Entering large log texts directly into Copilot Chat
Maintaining long conversations with GitHub Copilot
Having a moderate conversation that at some point contains a log paste
In all cases, you should be able to reduce your token usage by starting a new conversation. The reason longer conversations can cause issues is because Copilot will read the entire conversation history when considering a response, so starting fresh will often resolve this.
If you have cleared your conversation history but are still being regularly rate limited, please reach out and I will investigate this further for you."

I have had to reach out further, as I am still being rate limited after starting brand new chats, and making single line requests (even after several hours)...

[kumerian]

Yup, I guess we all got that boilerplate reply that doesn't really address anything. Next you'll be told that Roo is a problem even though you're not using it. That was the last thing I heard. Then it went all silent.

[cbunting99]

I too started getting the rate limit. But looking at this reply from the ticket, I'm not sure I understand why our tokens increase because of how long our conversations may be or because copilot rereads the conversation history. Tokens should only accumulate from writing text, code, edits etc.

[AwwCoconuts]

In my ongoing active ticket (2 weeks now) with Git support- they continuously just blame the IDE, suggest changing programs etc.

I opened a bug report with VSCode-Engineering, they have THOUSANDS of reports of this. Here's what has come of it:
The problem is that Git hub is selling this service as "UNLIMITED" pro subscription. Here's the crux of it; Every single model option, is TOKEN based usage from the AI provider themselves. ALL THE MODELS use Token amounts and history to rate limit (Claude is worst for this, even with private paid Claude account I get rate limited in their official app). Any code script that starts hitting 500 lines + will trigger these rate limits, period. Regardless of using co-pilot or the official provider accounts/apps.

This is not a Git issue. Where it becomes a Git issue, is that Git is who has sold us this "unlimited" access account that cannot and does not exist. Unless the providers of the AI themselves (Anthropic, etc.) enter some kind of truly "unlimited" token requests contract/agreement with GIT, the service is totally doomed.

Edit: Typos

[AwwCoconuts]

I was told by Git staff that when you submit any request, it also scans your entire chat history, edit history etc- WHICH USES TOKENS (yes, every request). Git's suggestion was to start a new chat or clear edit history every 10 minutes or so. (made zero difference for me)

The flaws in the service are in the actual design and implementation of the service agreements between providers. As the paying end users, we are stuck in the middle getting hosed as usual.

[khairu-aqsara]

I'm also facing the same issue. At this rate, what's the point of paying if I'm still hitting a rate limit? This isn't what they promised.

[Ehbraheem]

I just hit this right now. I think @github help explain what's going on

[nikushx]

They very clearly do not care it seems.

[nikushx]

Lost at least 10+ customers from my own circles alone. Confusing business model, guys!

[karrot0]

I discovered that not adding the files that I wanted to edit to the working set would just look for references and when there was too much code it would rate-limit me, so just add your working set files. Maybe that'll work

[rahul-s-bhatt]

I discovered that not adding the files that I wanted to edit to the working set would just look for references and when there was too much code it would rate-limit me, so just add your working set files. Maybe that'll work

I realized that too, thanks will try this method then.

[AwwCoconuts]

Sorry, your request failed. Please try again.

EVERY SINGLE TIME NOW.

[AwwCoconuts]

Lol... Today's nonsense: `````Python x 1500 lines

[Dayclone]

Try starting a new convo with it

[AwwCoconuts]

read. the above posts.... i specifically have a ticket with git support which already addresses. I start a new thread every 10 minutes. MAKES NO DIFFERENCE

[Dayclone]

I switched to a trial thinking I exceeded my limits but after upgraded I am still seeing the same UI at the bottom on my VS Code Insider. For my VS Code I can see the various LLM models. So something is not updating on VS Insider for me when using agent.

[hafieyz]

yes same with me..

[waleed1892]

I started liking github copilot with their latest agent mode but this rate limit is pissing me off. After every 5 minutes I'm rate limited.
This info should be mentioned in the pricing plan.

[Dayclone]

Try starting a new convo, and I also have it create a reference.md file and update that with a timestamp of what it's working on, problems found, solutions created, etc...

[OptimizeStock]

Nothing like waiting 56 minutes to use copilot. Is there a way around this as a paying user?

[Dayclone]

I find that if I start a new conversation it works, but sometimes I also get the rate limiting error. I've reached out to GitHub support and they can't really comment on this issue because it's a pre-release feature apparently. It's annoying.

[OptimizeStock]

Yes, unfortunately it didn't work, it looks like it's on a timer.

[OptimizeStock]

I have a feeling users are rate limit more often during peak hours. I was able to go for 5-6 hours before I was rate limited around 6 a.m.

[AwwCoconuts]

New day, open vs-code, 3 commands in using 01 preview hit a 7.5 hr rate limit. This was at 4 a.m. EST by the way...

[RBV801]

I just hopped on and immediately hit this issue. Pretty clear they are just taking our money. They won't fix it. This whole thread is Gits middle finger the users on it. So, cancel your subscription. I have no alternative, but just giving git money is also not an alternative.

[RBV801]

One amendment: you can just pay for the API more directly through a service such as OpenRouter, but then you are pay as you go.

At the end of the day there is a design issue with the overall AI pipeline: there is no clear feedback to the user on prompt efficiency. You may ask a simple question, but unknown to you (because there is no mechanism to show it) this triggers a huge load in the API call. The only current way to get this feedback, it seems to me, is to constantly change convos, but because there is no feedback the user has to do this arbitrarily, which really maximizes frustration and limits the ability to take advantage of the AI capabilities.

Unregulated solutions seem to address this from the UX POV, but the actual system doesn't work like that, enter the frustration with GitCopilot. Bottom-line, they have a lot of work to do before they can make the claims they make good on their claims for the subscription. Given this has been going on since the summer, they need to stop selling this product, but they won't until some decides it is worth it to sue (fat chance) or people just stop buying.

[Dayclone]

Use Roo Code

[hack-STACK]

Same here, im recently pay the subscription cause it worth to pay with, on the free plan i got Zero rate limit, so i pay to buy some coffe for the developer, but i get shock when i pay like every second i prompt it triggered the rate limited for generate, its kinda useless tho, and so ironic eventually.

[hack-STACK]

For the temporary solution was restarted your device or vscode, and prevent using an '#' (that gonna make way more chances the rate limiter trigger), or just sleep until the token being restart to default 😂.

[AwwCoconuts]

Going to bed / waiting the allotted time for penalty makes no difference. Waking up to a new day and entering 3 commands yields a 7.5 hr penalty. Changing model makes no difference. Service is broken. Period. My 3.5 week long ticket with Git support finalized with them suggesting using a different service that works. ffs. Couldn't be more of a joke

[ulbi]

Same problem here, pay for this stuff and run into undocumented and unclear rate limits after 10 prompts!

[ulbi]

After switching the model from o1-preview to gpt-4o problem is gone - very unclear communication, not transperancy at all - very bad.

[AwwCoconuts]

Switching model will work for a couple commands at best from my testing... Hope you get better results!

Again, as mentioned in the rest of the thread, the issue is with the API services the ACTUAL model providers use. ALL of them are token limited, GIT CANNOT provide an "unlimited" service without obtaining UNLIMITED tokens to/from the individual providers (such as Anthropic (which OBVIOUSLY they haven't.))

[ebynnerlys]

Tengo el plan de pago activado que supuestamente ofrece uso 'ilimitado' de Copilot (y a esperar que las ranas críen pelo, porque de ilimitado tiene lo que yo de astronauta 🐸).
Sobre los límites de Anthropic: Vamos a ver, si los límites los aplicara Anthropic, ¿cómo diantres saben los de OpenAI que he estado usando la API de Anthropic? Cada proveedor debería tener su propio contador, a menos que GitHub esté haciendo de chivato y compartiendo datos entre bambalinas.
He intentado cambiar de modelo como sugieres, pero ni por esas. Tu teoría sobre los límites de los proveedores tiene su lógica, no te lo niego, pero mis pruebas (con imágenes que lo demuestran) confirman que GitHub también mete la cuchara con sus propias restricciones.
Y la guinda del pastel es que esto pasa incluso en domingo, día de descanso donde los haya. ¿Dónde se puede dejar una reseña para expresar nuestro 'entusiasmo' por este servicio tan 'ilimitado'?