GitHub Community
Announcing GitHub Copilot Pro+ #155714
Replies: 51 comments 58 replies
-
|
I don't understand the Premium Requests table! I'm paying for GitHub Copilot Pro. I'm supposed to have 300 Premium Requests. Does that mean I can only use 300 Claude 3.7 Sonnet messages in agent mode/chat? Source: Model multipliers: (Claude 3.7 Sonnet has a value of 1 in the table). What is the base model? For a moment, I thought the base model was GPT-4o. |
Beta Was this translation helpful? Give feedback.
-
|
Please @Akash1134 could we get some sort of clarification on what premium requests means? The docs now are so confusing. What is the "base model"? Is it something proprietary or just 4o throttled? Can we get the math behind the "premium requests multiplier"? I use probably 100-150 requests per day. The way the documentation reads this makes it look like I'm going to have 2-3 days of premium requests and then I'm going to be paying $4/day after unless I want to use some model you guys won't even name that's worse than 4o. Which is a $100/mo price hike for business/pro users if they want to be productive. It needs clarification 🙏 |
Beta Was this translation helpful? Give feedback.
-
|
Here you go guys. For instance, if you stick with using claude 3.7 thinking model, it is not 300 chats per month, but 240 as it consumes 1.25 premium requests |
Beta Was this translation helpful? Give feedback.
-
|
Honestly, 300 requests are really too few. Moreover, you haven't provided information about the base model. |
Beta Was this translation helpful? Give feedback.
-
|
The base model is GPT4o. Check this table out |
Beta Was this translation helpful? Give feedback.
-
|
Base model is probably 4o-mini, need more clarification |
Beta Was this translation helpful? Give feedback.
-
|
Check this table out. It is GPT4o |
Beta Was this translation helpful? Give feedback.
-
|
so if i hit 300 premium request limit, the response fallback to shittier model no ? |
Beta Was this translation helpful? Give feedback.
-
|
I guess they will offer you to upgrade to Pro+ or to enable pay per use. The unlimited model is 4o, the one that often even refuse to answer for some reason... |
Beta Was this translation helpful? Give feedback.
-
|
I just subscribed to Pro+ one hour ago, and it keeps giving me the same error message.
|
Beta Was this translation helpful? Give feedback.
This comment was marked as spam.
This comment was marked as spam.
-
|
same here |
Beta Was this translation helpful? Give feedback.
-
|
Where did you even find how to upgrade to Pro+? I've been all over our org settings in github, including the billing areas, and it just shows me our existing GitHub Copilot Business option and nothing about upgrading to Pro+. |
Beta Was this translation helpful? Give feedback.
-
|
Have you figured this out? I am in the same position. Trying to change our orgs from Business to Pro+ |
Beta Was this translation helpful? Give feedback.
-
|
I was checking out Cursor's plans and it offers unlimited slow requests for Claude 3.7, so I think it's a better option than Copilot. 🤔 |
Beta Was this translation helpful? Give feedback.
-
|
its fake unlimited. Buy 1 month and try. Delay will quickly add to 2-3min per message |
Beta Was this translation helpful? Give feedback.
-
|
I see a lot of people say slow requests are typically only delayed by a few seconds at most. Some people report almost no delay at all, even. |
Beta Was this translation helpful? Give feedback.
-
|
It depends on the time of the day and the period of the year. I have used Cursor for a year, there was some months where the slow request were the same as the fast request, and other where the slow request were barely usable. You can't really depends on them. So that's why you will see different experience. Usually when it gets bad, you will see a lot of people complains on their forums about it. Their peak time is asian time for some reason. So if you use in North America timezone you will have a better experience in general on the slow request. But yeah, you can use a lot of them, it's really unlimited. |
Beta Was this translation helpful? Give feedback.
-
|
This has just made using copilot so over complicated and much more limiting. I might just move to cursor, especially if they'll be limiting use of ALL models except whatever 'base model' is. |
Beta Was this translation helpful? Give feedback.
-
|
So what is base model ??? Where can I find specific information about it? |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
However, according to the explanation, 4o is a premium request and will consume 1 request. |
Beta Was this translation helpful? Give feedback.
-
|
I wouldn't be surprised if it turns out to be GPT-3. "Unlimited requests!" How generous are you! |
Beta Was this translation helpful? Give feedback.
-
|
Any updates about this topic? What is a base model? |
Beta Was this translation helpful? Give feedback.
-
|
Check this out. It is GPT4o, the base model |
Beta Was this translation helpful? Give feedback.
-
|
According to this blog post, the base model is GPT-4o. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you so much for this! Honestly now I'm not that upset (assuming this is the current 4o, the version currently beating 3.7 sonnet) |
Beta Was this translation helpful? Give feedback.
-
According to the table provided here, "Base model" and "GPT-4o" are listed as separate rows, which really confuses me. |
Beta Was this translation helpful? Give feedback.
-
Now the table is updated. GPT-4o is the base model
|
Beta Was this translation helpful? Give feedback.
This comment was marked as spam.
This comment was marked as spam.
-
|
My guess is they still offer unlimited request until May 5th, so they will probably remove the rate limit after that date. |
Beta Was this translation helpful? Give feedback.
-
|
I’m using Cursor and Windsurf for iOS development, but I’ve faced several issues—especially with Synthetics and compiler not building properly. However, the Copilot Agent performs much better. It understands context more accurately and introduces fewer bugs or issues. That said, I hit the access limit after just one hour, which is really strange. |
Beta Was this translation helpful? Give feedback.
-
|
Current limits are per hour and per day |
Beta Was this translation helpful? Give feedback.
-
|
It’s disappointing |
Beta Was this translation helpful? Give feedback.
-
|
"We're excited to introduce GitHub Copilot Pro+" Wow, so exciting, we now need to pay more for less. @Akash1134 Can you confirm that those of us who are on a yearly plan won't be affected until our next renewal? Since, you know, that's really not what we signed up for? |
Beta Was this translation helpful? Give feedback.
-
|
From what little information I’ve found or read there seems to be no grandfathering and everyone including those with yearly subscriptions will be switched over to the new limits starting in May |
Beta Was this translation helpful? Give feedback.
-
|
Can't wait for Jules from google to come out.. hopefully adding some big boy competition against github; curious to see if they also adopt monthly requests limits - although knowing their existing API limits - it's likely to be liberal - keeping AI studio API rate limits (minute/hour) vs monthly boundaries.. I believe it will also be a coding sft/RL specialized model. This pricing change smells like a desperate cash grab from GitHub - especially given how vague they are about the details. MS marketing strategy was to lure user in with liberal API usage; classic 'bait and switch" |
Beta Was this translation helpful? Give feedback.
-
|
I have reviewed the GitHub Copilot documentation and performed some straightforward calculations, and quite frankly, the results are extremely disappointing. According to the documentation, using Claude 3.7 Sonnet Thinking is rated at 1.25 units per request. If I'm a Copilot Pro user with a quota of 300 units per month, that translates to 240 requests total. Dividing that by 30 days, it means I'm limited to only 8 requests per day. This feels less like a premium AI service and more like an aggressively limited trial — especially considering the subscription is marketed as a "Pro" tier. I understand there are costs involved in providing high-quality models, but the current limits feel severely misaligned with both expectations and the pricing. I urge the Copilot team to seriously reconsider this quota model. We're willing to pay for quality — but we also expect fairness. I also purchased JetBrains AI Assistant as a comparison. While its capabilities may lag behind in some areas, at least the model usage is not restricted by hard limits. Their approach appears to be dynamically balanced, rather than relying on a rigid quota — and certainly doesn't result in an awkward situation where users are limited to just 8 requests per day. Considering that Copilot's integration with JetBrains products is still less seamless than that of JetBrains' own assistant, this quota model may actually push power users to switch to JetBrains AI Assistant. And let's not forget other potential competitors like Cursor, who are rapidly gaining ground. |
Beta Was this translation helpful? Give feedback.
-
|
So, regardless if you have a pro sub already or not, the price for pro+ is the same? |
Beta Was this translation helpful? Give feedback.
-
|
o3 isn't even coming to Copilot Pro. They genuinely do not care. Cancelling now :/ |
Beta Was this translation helpful? Give feedback.
-
|
So it looks like GitHub is simply ignoring the pricing hate. Very bad move, they're going to lose a lot of people. |
Beta Was this translation helpful? Give feedback.
-
|
I’ve used Copilot for over a year. At this price point, I think it’s time to switch to JetBrains. My main development tool is JetBrains anyway—I only chose Copilot because it worked across multiple tools. But now, with this pricing… well, goodbye |
Beta Was this translation helpful? Give feedback.
-
|
With all due respect, Copilot is currently prone to failure due to network issues, which often results in high-tier requests being deducted without any data returned. Has the team considered this? Therefore, limiting access to advanced models seems to be meaningless. |
Beta Was this translation helpful? Give feedback.
-
|
Limiting the advanced model to 300 requests per month feels like a joke. Only 10 requests per day? Which paying user could possibly accept that? |
Beta Was this translation helpful? Give feedback.
-
|
and now the autocomplete is worthless lmao |
Beta Was this translation helpful? Give feedback.
-
|
The recent slowdown of Copilot Pro is unbearable. Is this a tactic to convert us users to Copilot Pro+? If you play such a disgusting game, who will stay? |
Beta Was this translation helpful? Give feedback.
-
|
also the models seem stupider, compared to cursor this is unjusticed action 😭😭 |
Beta Was this translation helpful? Give feedback.
-
|
I guess they choose the worst path to kill the PRO subscription, instead of just rise the price. |
Beta Was this translation helpful? Give feedback.
-
|
Hey GitHubians! I'm writing this because the new request limits are honestly pretty frustrating. Moving from unlimited requests to only 300 requests per month for the Pro plan and 1,500 requests for the Pro+ plan is completely unreasonable. I work 15 hours a day and use an average of 10 requests per hour, which is not excessive given that each request typically takes about one minute to complete. That amounts to only 10 minutes of usage per hour. This results in 150 requests per day, totaling 4,500 requests per month. However, with Claude 3.7 Sonnet Thinking, each request is multiplied by 1.25, bringing the monthly total to 5,625 premium requests. The Pro+ plan at $39/month only offers 1,500 premium requests, which is far from sufficient. To cover my needs, I would have to either:
This is a massive increase from the previous $10/month, representing a 1,940% price hike. I get that pricing needs to be adjusted, but these new plans just don't make sense for those of us who use Copilot intensively. I strongly encourage you to consider more flexible plans or options for high-volume users. Thanks for reading! Best regards, |
Beta Was this translation helpful? Give feedback.
-
|
Rest assured, there already was a limit. They were simply lieing about it. There's a 100+ pages thread on the forum about people getting errored out with "you hit the rate limit" errors although it looks like they deleted that thread. Move to Cursor. Way better results anyway. |
Beta Was this translation helpful? Give feedback.
-
|
What you're saying isn't entirely accurate. Yes, there was a rate limit that I hit several times, but it didn't last very long — usually 30 minutes to an hour max. Plus, that limit wasn't applied globally to all models. For example, if I exceeded the quota on Claude 3.7 Sonnet Thinking, I could still switch to Claude 3.7 Normal, and it would work. Now, they're clearly talking about a limit based on the number of requests per month, which is very different and much more restrictive. And just to add, there's another form of limitation I've noticed. When using the API or via Cursor, I've found that the request processing speed is noticeably faster than when the requests come from GitHub Copilot, even for the same tasks. So, to summarize: Yes, there was a rate limit, but it wasn't based on a monthly quota. And no, it wasn't as restrictive as it is now. |
Beta Was this translation helpful? Give feedback.
-
|
At least now they're honest about there being a rate limit. There always was one, judging from the thread on the forums where hundreds of people complain about running into the rate limit within minutes each day and being told "come back tomorrow". The thread has mysteriously vanished, the problem hasn't. I've moved on. Github Copilot has been surpassed by the competition. Both in features and in pricing. |
Beta Was this translation helpful? Give feedback.
-
|
I agree, they weren't totally transparent before, but what they're doing now is even worse. Before, the limit was temporary and lasted a max of 1 hour, and you could switch models to keep working. As for Cursor, yeah, it's great, but 500 requests per month is way too low for heavy users. In 3-4 days, I've already burned through all of it. Plus, they also have their own request quotas. Honestly, it's getting harder to find a reliable solution. I'm considering Google Code Assists, but their interface is seriously terrible. Maybe with a good reverse proxy, it could work… |
Beta Was this translation helpful? Give feedback.
-
|
Feels like soon we’ll be watching ads before every request |
Beta Was this translation helpful? Give feedback.
-
|
So let me get this straight. Previously, you'd get unlimited amount of requests for Claude 3.7 and other advanced models, but with these changes, the Pro plan is only limited to 300 requests per month even for paying customers? You know what, I intentionally try to limit my usage and reliance on AI when I am developing, so while I haven't counted how much I use Claude in the past, I sincerely doubt it ever exceeded this 300 request/monthly quota. But having the peace of mind that I can use it freely is what made Copilot Plus a good deal for me. Right now with these introduced changes, it feels like this is something I'd have to constantly have to worry about and keep in the back of my mind, so god forbid I don't hit some imaginary limit. It would have been infinitely better to have some sort of soft limit, and once a customer exceeds that limit you can start rate-limiting or throttling their requests. The option you went with instead just shouts pure greed, and I've been happily using Copilot (even with all it's flaws) for over a year, but as soon as I see a message I've exceeded some quota asking me to pay more, I'm cancelling my plan on the spot.
|
Beta Was this translation helpful? Give feedback.
-
|
they just disabled the audit logs from the copilot, no wonder my agents not CANNOT read the terminal log! i am done now, cancelling right now.
|
Beta Was this translation helpful? Give feedback.
-
|
I think you misunderstood Copilot. The audit logs are the logs of your actions on your GitHub management page. The logs in your terminal are related to tool calls. Copilot would never restrict the ability to read VS Code terminal logs, because it would be meaningless to do so. |
Beta Was this translation helpful? Give feedback.
-
|
I dont know how i missed this, but premium requests arent enforced yet. |
Beta Was this translation helpful? Give feedback.
-
|
I upgraded to Copilot Pro+, but I do not see o3 and 4.5 models in chat in the VS Code and on web site. How to enable them? Tried to restart VS Code. |
Beta Was this translation helpful? Give feedback.
-
|
I enabled them manually. |
Beta Was this translation helpful? Give feedback.
-
|
First we didnt get o3, now we're not getting Claude 4 Opus. Insane. |
Beta Was this translation helpful? Give feedback.
-
|
To be honest, at present, the number of APIs given is very small, and I am also trying to switch to other platforms |
Beta Was this translation helpful? Give feedback.
and others
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
We're excited to introduce GitHub Copilot Pro+, a new tier designed for developers eager to elevate their coding experience. Building upon the robust features of GitHub Copilot Pro, this enhanced plan offers exclusive benefits to supercharge your development workflow.
What's New in GitHub Copilot Pro+
To get started, purchase GitHub Copilot Pro+ today and explore the additional Copilot updates highlighted in our announcement blog.
Stay tuned as we continue to enhance your developer experience. Please share your feedback in the comments below. Happy coding! 🚀
Documentation
Beta Was this translation helpful? Give feedback.
All reactions