Selected Claude, OpenAI, and Gemini Copilot models are now deprecated

[KateCatlin]

Today we deprecated 8 models across Claude, GPT, and Gemini families to streamline our model catalog. These models have been superseded by newer versions that offer improved performance and capabilities.

This follows a warning 30 days ago that these models would be deprecated.

---

Models Being Deprecated

Model	Provider	Suggested alternative
Claude Sonnet 3.7	Anthropic	Claude Sonnet 4.5
Claude Sonnet 3.7 Thinking	Anthropic	Claude Sonnet 4.5
Claude Opus 4	Anthropic	Claude Opus 4.1
GPT o3	OpenAI	GPT-5
GPT o1 mini	OpenAI	GPT-5 mini
GPT o3 mini	OpenAI	GPT-5 mini
GPT o4 mini	OpenAI	GPT-5 mini
Gemini 2.0 Flash	Google	Gemini 2.5

---

What to Expect

• Deprecated models will be removed from model selection interfaces
• If a deprecated model is your current default, an alternative from your available models will be automatically selected
• Requests to deprecated models will return a 400 error

---

Questions?

Please share your feedback and questions in this discussion post. If you have concerns about any of these deprecations or need guidance on migrating to newer models, we're here to help!

[ShreyaPandayy]

Thanks for the update! 🚀

It’s good to see the catalog being streamlined, but I’d like to share some feedback:

Could you please provide clear migration guides (code snippets, API request examples) for developers who were using the deprecated models (like gpt-o3-mini → gpt-5-mini)?

Will there be any pricing or quota changes when we are moved to the suggested alternatives?

For teams running production workloads, a longer deprecation notice period or automatic upgrade tool would be very helpful to reduce migration friction.

Overall, the upgrade makes sense, but a smoother migration path and more detailed docs would make the transition much easier. Thanks for keeping the ecosystem evolving! 🙌

[KateCatlin]

Thank you @ShreyaPandayy for the feedback! We're actively iterating on this process so really appreciate hearing from the community on this.

As a quick answer, you can see more about multipliers per model here!

For your other suggestions, I think I need a better sense of your setup and workflow. Can you share more about:

• Which product or integration are you using these deprecated models with? (e.g., Copilot in your IDE, Copilot Extensions, etc.)
• When you mention "production workloads," can you give us a sense of what that looks like on your end?

This will help us make sure we're addressing your specific needs. Thanks again in advance!

[ShreyaPandayy]

Hey! @KateCatlin ⭐

Thanks for your response and for looking into this further. I'm happy to provide more context about our setup and workflow:

Current Integration Usage:

· Primary: Copilot in VS Code for daily development work
· Secondary: GitHub Copilot Chat for code explanation and debugging
· Experimental: Testing Copilot Extensions for specific frameworks (React, Node.js)

Production Workload Context:
Our team uses these models primarily for:

· Code generation for routine components and API endpoints
· Code review assistance - analyzing pull requests and suggesting improvements
· Documentation generation for internal APIs and codebases
· Bug detection and troubleshooting in existing codebases
· Test case generation for our quality assurance pipeline

The "production workloads" specifically refer to how these AI-assisted outputs get integrated into our development lifecycle. For instance, generated code snippets often serve as starting points that our team then refines and integrates into our main codebase. The deprecation affects our established workflows, particularly around code consistency and team onboarding.

What would be most helpful for us is:

1. Clear migration paths with comparable capability models
2. Advance notice timelines for future deprecations (90+ days ideally)
3. Documentation on feature parity between deprecated and replacement models

Happy to provide more specific details if that would be helpful. Thanks again for your attention to this!

[naina933]

A lot of folks still love Sonnet 3.7 (see this discussion: https://github.com/orgs/community/discussions/177463).

Unless there's a hard blocker on your side, could you temporarily bring Sonnet 3.7 back to the Copilot subscription plan for the next ~120 days? We've got a ton to do—update prompt strategies, plan the 3.7 -> 4.x migration, etc. Sonnet 3.7 and 4.x behave quite differently, so they need different prompting approaches, and a full migration in a day or two just isn't realistic.

This would really help keep our team afloat—it's basically just me handling prompt engineering.

We need that buffer to:

• Run targeted prompt research for 4.x
• Rework system prompts to match 4.x behavior
• Pilot in production and A/B test before switching defaults
• Document changes and train the team (because Claude 4.x is sensitive to how user input is phrased)

A short extension for 3.7 would make a huge difference. Thanks for considering it!

[Jinz426]

we should develope gpt 8 now