OpenAI has now retired GPT-5.1 in ChatGPT and auto-migrated chats to newer defaults, while keeping API availability unchanged.
What changed
- OpenAI’s Help Center states GPT-5.1 was retired in ChatGPT as of March 11, 2026.
- Existing GPT-5.1 conversations are now redirected to newer defaults: GPT-5.3 Instant for standard chats, and GPT-5.4 Thinking or GPT-5.4 Pro for deeper reasoning paths.
- OpenAI says GPT-5.1 remains available in the API, so backend integrations can continue without forced migration.
Why it matters for product teams
- Chat interfaces are becoming more opinionated. End users get upgraded models by default, reducing manual model-switch friction.
- Product reliability improves when model tiering is explicit: "Instant" for speed, "Thinking/Pro" for depth.
- Teams building consumer AI UX should expect model retirement cycles inside hosted chat products to be faster than API deprecation cycles.
- This split creates a clear strategy: keep API model pinning for production stability, while allowing app-level chat defaults to evolve faster.
Practical takeaway
If your workflow depends on ChatGPT model behavior, revalidate prompts, output style, and latency assumptions after this retirement change. If you rely on the API, you can keep GPT-5.1 for now, but plan migration testing to newer GPT-5.x variants before future retirement windows tighten.
Primary sources
- OpenAI Help Center: https://help.openai.com/en/articles/11484097-gpt-5-1-in-chatgpt-retirement
- OpenAI Help Center (ChatGPT release notes, March 11 section): https://help.openai.com/en/articles/6825453-chatgpt-rlease-notes