GitHub Community
OpenAI GPT-5 and GPT-5 mini are now generally available in GitHub Copilot #168107
Replies: 38 comments 62 replies
|
How long will it take GPT-5 to show up in VS Code after it's enabled in the organization settings? |
|
Some(many times) you need reload. |
|
The waiting has only begun at this point. The chatgpt5.0 agent is pathetically slow. Careful what you wish for :) |
|
many times need to reload. |
|
This is great news! 🎉 I’ve been waiting for GPT-5 integration in Copilot, especially for its improved reasoning and faster code generation. I just enabled GPT-5 mini on my free plan, and it already feels smoother. Planning to try the full GPT-5 soon to see how much of a difference it makes in complex workflows. |
|
Wait a minute...is it advanced reasoning or improved reasoning? When one considers advanced reasoning in these roaring ai 20's then comparison to U.S.'s chemical laser defense technology, compared to other countries and laughable... |
|
Will it replace 4.1 as the default model? |
|
I hope there are plans to make it the base model, since the cost is so similar to 4.1. Charging 1× premium request credits doesn’t seem right, given that Sonnet 4 is much more expensive than GPT-5, yet they have the same premium request cost. |
|
They should do it now since the competition is giving a free week of 5.0 |
|
Aqui não estava aparecendo, então fui nas configurações do GitHub Copilot, desativei todos os modelos e reiniciei o VS Code. |
|
I found on this page: https://docs.github.com/en/copilot/reference/ai-models/model-hosting
|
This worked for me thanks! |
|
I can’t see it in my VSCode, but some people can. |
|
Update your Vscode |
|
yes possibly its available for some of accounts |
|
According to their release notes, they plan to deprecate all models except GTP-5. Does this mean the new "default" model becomes GPT-5 for copilot? |
|
THis is in Chatgpt, not the API... |
|
i'm just a personal account but I had to go here and enable GPT-5 https://github.com/settings/copilot/features |
GPT-5 is enabled but not showing in VsCode. I'm on the latest version of VsCode with everything updated.
Version: 1.103.0 (system setup) |
|
Aqui não estava aparecendo, então fui nas configurações do GitHub Copilot, desativei todos os modelos e reiniciei o VS Code. |
|
I did the same |
|
The model consumes one premium request, right? If so, that's quite weird, since the price is even lower that GPT 4.1. |
|
When I subscribed to GitHub Copilot Pro for a year, there was no Pro+ tier. Now, GPT-5 is available to Pro users but consumes premium credits (multiplier 1), while GPT-4o doesn’t. Considering OpenAI charges GPT-5 at $0.625 per 1M tokens for input and $5 for output—compared to $3.75 and $15 for GPT-4o—GPT-5 is actually cheaper to run than GPT-4o. Yet GitHub treats GPT-5 as a premium feature. This feels less like a technical necessity and more like a monetization strategy. The introduction of Pro+ and credit-based restrictions makes the whole system feel like a milking cow rather than a good value-for-money solution. It’s pushing me away from renewing. 🔍 API Cost Comparison: OpenAI vs GitHub Copilot
Key Takeaways:
|
|
I totally agree with you! |
|
Maybe GPT-4o was subsidized as a loss leader to Microsoft and GPT-5 ends that policy. But having a free model which you can use when your premium credits run out for the month is extremely practical, so hopefully they keep 4.1/4o around for a long time, or introduce another 0-credit option later. |
|
I'm using GPT-5 (Preview) in Copilot in VS Code. When providing it context by dragging/dropping a folder of code to the chat window then offering it a prompt, the folder shows as Context but the response I get from the model gives me general feedback and indicates if it has the code it could be more specific. Then I try again to drag the folder in, ask it if it can see the context, it says Yes, I can see...rattles off some files and functions, and says "Tell me what you want to change or debug". I then try a third time, copy/paste my prompt from earlier into the same chat and explicitately tell it to use the folder (by folder name) as context and it then seems to do so. Previous models did not need to be reminded to use provided context and could keep a train of thought from message to message. |
|
You got access how? |
|
When I went into VS Code, there was an update for the Copilot extension. When I ran it and it restarted VS Code and GPT-5 (Preview) was available under Ask in the chat panel. I'm a Pro+ user. Not sure if that is a factor or not. I'm also a very heavy user. And keep a high budget for additional Premium Requests. Again, not sure if that factors into rollout. |
Same pro+ here. But unavailable. |
|
Probably rolling it out slowly. It's crawling right now. Given the launch buzz, I was doing some blind testing (just giving it a semi-detailed prompt and running with what ever code it spit out in an isolated sandbox account for this purpose). It's failed miserably so far. Nothing but error after error. |
|
The reasoning will be low, medium or high? |
|
I'm surprised that this is not specified ANYWHERE. |
|
im looking for that aswell, but asking the gpt gives this result |
|
For good reference, non-following. At start of a column section transformation. However, of B-Spline expression polar forms is far ahead of the above options. In that area options first, outer, inner and last terms can all be used in multiple imaginary axis. |
|
I can see GPT-5 on Github online but not in VsCode with the extension enabled. Anybody knows why ? |
|
1 premium request? lol |
|
@Lolo401 - it is literally listed in your screenshot... ? |
Commenter said the following, which leads me to believe the screenshot is from a vscode.dev instance.
|
|
I've been testing various models in Copilot and noticed they're all capping out at around 128k context length (Found this out with some debugging), even though some like GPT-5 are supposed to handle 400k. This is causing conversations to get summarized way too early and breaking continuity. Has anyone else run into this? Is this a known limitation right now, or am I missing something in the settings? Really hoping this gets bumped up to the full supported lengths soon - would make such a difference for longer conversations and complex tasks.
|
|
I hope new Claude models which have 1M context also to be added soon, those will be game changer! |
|
When will ChatGPT 5 be available in the Visual Studio 2022 Extension? |
|
Soon! We're testing GPT-5 in Visual Studio at the moment and hope to make it available to all soon in days not weeks. Disclosure: I work on the Visual Studio Copilot team. |
|
Why does a model over an order of magnitude cheaper than Claude 4 and comparable in price to unlimited GPT 4.1 cost entire one premium request, especially when it's in preview? |
|
It'll probably come eventually- it is brand new. Hosting of models for Copilot Chat
|
Because this whole GitHub Copilot thing is becoming a milking cow, that’s why…See also my other reaction with the comparison of both models and the fact it should never be a premium model: When I subscribed to GitHub Copilot Pro for a year, there was no Pro+ tier. Now, GPT-5 is available to Pro users but consumes premium credits (multiplier 1), while GPT-4o doesn’t. Considering OpenAI charges GPT-5 at $0.625 per 1M tokens for input and $5 for output—compared to $3.75 and $15 for GPT-4o—GPT-5 is actually cheaper to run than GPT-4o. Yet GitHub treats GPT-5 as a premium feature. This feels less like a technical necessity and more like a monetization strategy. The introduction of Pro+ and credit-based restrictions makes the whole system feel like a milking cow rather than a good value-for-money solution. It’s pushing me away from renewing. 🔍 API Cost Comparison: OpenAI vs GitHub Copilot
Key Takeaways:
|
|
If you're upset with the pricing, buy your own API key and use the custom API key option. When you realize exactly how much tokens the Agent mode consumes that you're not being charged for you will be a lot more appreciative of the situation. Even for casual use, you would easily exceed $10-40 dollars a month just on token usage and that's even with using GPT-5 nano. |
I already have my own API keys that I use for various integrations beyond GitHub Copilot, so purchasing them isn't the issue. I understand the token costs in Agent mode are high and currently subsidized, but my concern is the shift where even lower-priced models are gated behind premium tiers. This suggests a future where users need both a subscription and additional credits to access basic features, potentially pricing out casual users without clear value adds. |
|
I think it would be better to replace GPT 4.1 with GPT 5 because it's 1 entire premium request for something that shouldn't be |
|
It'll probably come eventually - it is brand new. Hosting of models for Copilot Chat GPT-4.1 is hosted by GitHub's Azure tenant when used in GitHub Copilot. |
This comment was marked as off-topic.
This comment was marked as off-topic.
|
When I tried to run GPT-5 on copilot as Agent, it's not showing any conversation with any verbosity, just straight up code and tools iterations. Is there a setting needed to be configured or a mistake on rolling out the model? 
|
|
same experience. it's DISGUSTING! |
|
One thing I'm pretty sure it is not disgusting, in actuality this model handles a niche that sonnet or opus can't on my side. So beside the lack of verbosity and some interaction with the vscode as a whole like running the scripts or debugging to get contexts like what sonnet/opus does, it's pretty nice. |
|
GPT 5 is absolutely HORRIBLE in Github copilot pro. It gives extremely weak summaries or explanations of what it's doing. It just makes changes and poorly and gives a terrible explanation of what it does. Extremely underwhelming and disappointing. Anyone else have this experience? So far, Claude Sonnet 4 is FAR superior. |
|
Yes very bad in terms of explanation and never follow my described tasks if more than one just ignore it. Very naughty model. |
|
Folks, charging 1x premium request for a model which is cheaper than gpt-4.1 doesn’t feel right at all |
|
Hi folks, any idea when GPT5 will be available in General Availability (GA)? |
Introducing GPT-5 mini: Now Available in GitHub Copilot (Public Preview)! 🚢 ✅ |
|
We cannot thank you enough for taking the community's opinions into consideration and making this breakthrough. |
|
Awesome, but a lot think that GPT-5 should either be free for paid plans, as it is wayy cheaper than 4o or 4.1. Or make it at least cheaper, because one entire premium request is ridiculous. |
This comment was marked as off-topic.
This comment was marked as off-topic.
|
I think so gpt-5 - best for complex, multi-step reasoning and high-quality code generation and gpt-5 mini - best for fast, lightweight coding tasks with lower latency. We are talking about power vs speed. |
whay is the reason of this, isn't it zero cost model? why is the problem? |
|
same here lol. this is a joke .. seriously |
|
This is the slowest model I have ever used, and is the least like to produce usable results, follow directions accurately, or even make sense of moderate to high complexity requests. Very unimpressed. |
As an Asia user, I’ve found that sometimes GPT5 works amazingly and quickly. Perhaps I avoid the peak workload times. |
|
My first time using this program I will update thank you for your feed back
Yahoo Mail: Search, Organize, Conquer
On Sat, Aug 16, 2025 at 1:06 AM, ***@***.***> wrote:
This is the slowest model I have ever used, and is the least like to produce usable results, follow directions accurately, or even make sense of moderate to high complexity requests. Very unimpressed.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>
|
|
OpenAI's GPT-5 and GPT-5 mini are now generally available in GitHub Copilot, bringing advanced reasoning, high‑quality code generation, and fast, cost‑efficient assistance to more developers. Availability in GitHub Copilot
You can access both models in GitHub Copilot Chat via the model picker on:
|
|
Once in an hour or so after using GPT-5 I get this message: |
|
GPT-5 and GPT-5 Mini Now Available in GitHub Copilot OpenAI’s GPT-5 and GPT-5 mini are now generally available in GitHub Copilot, bringing stronger reasoning, higher-quality code generation, and faster, more cost-efficient assistance to developers. Availability
Both models can be accessed via the model picker in Copilot Chat across:
How to Enable Access For Copilot Enterprise and Business
For Individual Paid Plans
Feedback Users are encouraged to share feedback to help improve model performance and developer experience. For more details, refer to GitHub Copilot’s official documentation on available models. GPT-5 brings deeper reasoning, while GPT-5 mini balances speed and cost—giving developers flexibility depending on the task at hand. |
Use the attached folder "<FOLDER_NAME>" as the primary and authoritative source. If something is unclear, say exactly which file is missing instead of guessing. Task: Requirements:
Issue: Task: Constraints:
Focus on:
Re-run the analysis using "<FOLDER_NAME>" and reference specific files/functions. |
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.
OpenAI's GPT-5 and GPT-5 mini are now generally available in GitHub Copilot, bringing advanced reasoning, high‑quality code generation, and fast, cost‑efficient assistance to more developers.
Availability in GitHub Copilot
You can access both models in GitHub Copilot Chat via the model picker on:
Enabling access
For Copilot Enterprise and Business: Administrators must opt in by enabling the new GPT-5 and GPT-5 mini policy in Copilot settings. Once enabled, users in that organization will see GPT-5 mini in the model picker.
For individual paid plans: Enable either model by selecting it in the Visual Studio Code model picker and confirming the one-time prompt.
Share your feedback
Join us in the comments below to share your feedback. To learn more about the models available in Copilot, see our documentation on models and get started with Copilot today.
All reactions