OpenAI’s latest model, o3-mini, now available in GitHub Copilot and GitHub Models

[github-staff]

OpenAI’s latest model, o3-mini, is available in GitHub Copilot and GitHub Models, bringing even more advanced AI capabilities to your coding workflow.

The o3-mini reasoning model surpasses o1 on coding benchmarks and matches o1-mini in response times, delivering improved quality with nearly the same latency.

Availability

This advanced model is now available to GitHub Copilot Pro, Business, and Enterprise users

Access it via the model picker in Visual Studio Code or GitHub chat. Support for DotCom, Visual Studio, JetBrains is coming soon!

How to Use

To start enhancing your workflow—whether debugging, refactoring, modernizing, testing, or more—simply select "o3-mini (Preview)". Paid Copilot subscribers get up to 50 messages every 12 hours.

Business/Enterprise admins can enable access for org members through their admin settings here.

For GitHub Models Users

GitHub Models users can also utilize the o3-mini model to boost their AI applications and projects. Available through the GitHub Models playground, o3-mini allows comparisons with other top models like:

• DeepSeek-R1
• Llama-3.3-70B-Instruct
• OpenAI o1-mini

Experiment with sample prompts, refine your ideas, and iterate to build powerful AI-driven solutions. Explore o3-mini’s versatility alongside other leading-edge models!

To learn more about, check out product documentation on GitHub Models.

Got feedback or questions? Share them in the comments below. 🚀

[iwangbowen]

It's very useful

[sharf666]

What's the reasoning_effort level used by Copilot?

[nickchomey]

im also curious about this

[truong211]

Medium

[TharunMJ]

Would you like to learn how to customize Copilot’s reasoning depth to better suit your programming needs?

[ghzgod]

Is this o3-mini as in the base model or o3-mini-high (the model for coding) which would make more sense for copilot

[ericktreichel]

Very relevant question, that's something we would love to know! If its mini-low, I would prefer keep Claude 3.5 Sonet, if its mini-high, I would prefer to use it in order to confirm if its better than Sonet 3.5.

[hmorneau]

It's very low, probably lower than low. It barely answer. You can give it a try, you will see for yourself.

o1-mini was answering with 1500 to 2000 tokens.
o3-mini output at most 200 tokens. #4530

Same problem as for o1 #3505

[eevmanu]

https://github.blog/changelog/2025-02-06-openai-o3-mini-is-now-available-in-github-copilot-free/

The model is configured to use OpenAI’s medium reasoning effort

[rafek1241]

Waiting for that in jetbrains IDE! :)

[abitskil]

It will be available in a year. =)

[RyuX51]

Me as well. I'm paying Copilot Pro since the beginning and still can only 4o in the IDE. I will not extend my subscription im May if it's still not possible to use a reasoning model there.

/edit:

I just got this roadmap link from the support:

https://github.com/orgs/github/projects/4247/views/1?filterQuery=model+picker

So it's planned for Q1.

[abitskil]

Yeah, I get that VS Code is the priority, but JetBrains still has features that work much better!

[bjontario]

New o3-mini is not available in compare dropdown ?

[rrva-mc]

What about the ability to select model in the plugin for Jetbrains IDE:s? This feature is missing from the Github Copilot plugin for Jetbrains and is starting to make alternatives for that family of IDEs look better than the stock GitHub plugin, which is a shame.

[RyuX51]

I already answered that in another question. It's planned for Q1: https://github.com/orgs/github/projects/4247/views/1?filterQuery=model+picker