← Back
OpenAI
OpenAI releases GPT-5.4 mini in ChatGPT; adds mid-response adjustments and improved reasoning
OpenAI APIChatGPT · releasefeaturemodelapi · help.openai.com ↗

GPT-5.4 Mini and Thinking Models

OpenAI has begun rolling out GPT-5.4 mini and GPT-5.4 Thinking to ChatGPT. GPT-5.4 mini is available to Free and Go users via the "Thinking" feature, and serves as a rate-limit fallback for paid users when GPT-5.4 Thinking reaches capacity. For Enterprise customers, GPT-5.4 mini can be set as the default model in Auto routing if preferred. The model will not appear as a selectable option in the model picker.

Key Improvements in GPT-5.4 Thinking

GPT-5.4 Thinking combines advances in reasoning, coding (incorporating industry-leading capabilities from GPT-5.3-Codex), and agentic workflows. Notable improvements include:

  • Mid-response adjustments: Users can now see an upfront plan of the model's thinking and adjust course while it's working, reducing back-and-forth iterations
  • Enhanced web research: Particularly improved for highly specific queries with better context retention
  • Improved context window management: Enables longer thinking sessions while maintaining relevance
  • Faster delivery: Complex work is accomplished more accurately and efficiently

Model Updates and Retirements

OpenAI is retiring legacy models and introducing refinements across its lineup:

  • GPT-5.3 Instant updates (March 16): Reduced teaser-style phrasing ("If you want…", "You'll never believe…") for more natural conversations
  • GPT-5.3 Instant updates (March 3): Improved accuracy, richer web search results, and better conversational flow
  • GPT-5.2 Instant personality update (January): More conversational tone with better contextual adaptation
  • GPT-5.1 retirement (March 11): GPT-5.1 Instant, Thinking, and Pro are no longer available; existing conversations automatically migrate to current equivalents (GPT-5.3 Instant, GPT-5.4 Thinking, GPT-5.4 Pro)
  • Legacy model retirement (February 13): GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini removed from ChatGPT

Under-18 Safety Principles

OpenAI updated its Model Spec in December 2025 with new Under-18 (U18) Principles that add age-appropriate guidance for users aged 13-17, building on existing safety rules while recognizing developmental needs of teen users.