← Back
OpenAI
OpenAI rolls out GPT-5.4 mini in ChatGPT; retires older GPT-5.1 models

GPT-5.4 Mini Rollout

OpenAI has begun rolling out GPT-5.4 mini across ChatGPT, available to Free and Go users via the "Thinking" feature. For Plus, Pro, and other paid users, GPT-5.4 mini serves as a rate-limit fallback for GPT-5.4 Thinking, ensuring continued access to reasoning capabilities during peak usage periods. Enterprise customers retain the option to default Auto routing to GPT-5.4 mini if preferred.

The new model will not appear as a selectable option in the model picker. Notably, GPT-5 Thinking mini will be retired in 30 days as a selectable model.

Model Retirements

As of March 11, 2026, GPT-5.1 models are no longer available in ChatGPT, including GPT-5.1 Instant, GPT-5.1 Thinking, and GPT-5.1 Pro. Existing conversations using these models automatically migrate to corresponding current versions: GPT-5.3 Instant, GPT-5.4 Thinking, or GPT-5.4 Pro.

GPT-5.4 Thinking Improvements

GPT-5.4 Thinking now includes several key enhancements:

  • Mid-response course correction: Users can adjust the model's direction while it's actively working through a response
  • Improved deep web research for highly specific queries
  • Better context maintenance for longer thinking sequences
  • Enhanced context window management supporting extended reasoning
  • Upfront thinking plans that clarify the model's approach before final output

These improvements enable faster delivery of higher-quality answers while maintaining relevance to the task at hand.

GPT-5.3 Instant Updates

The latest version of GPT-5.3 Instant includes refinements to tone and conversational quality:

  • Reduced teaser-style phrasing (e.g., "If you want…", "You'll never believe…")
  • More measured, grounded tone appropriate to conversation context
  • Improved web search results with better contextualization
  • Clearer, more relevant answers for advice-seeking and how-to questions
  • Important information placed upfront

Developers and API users should review the OpenAI blog post for additional details on these changes and their implications for existing applications.