GPT-5.3 Instant Tone Improvements

OpenAI has released an update to GPT-5.3 Instant that enhances response quality by improving follow-up tone and eliminating teaser-style phrasing. The model now reduces common conversational patterns like "If you want…", "You'll never believe…", and "I can tell you these three things that…" which can feel promotional or artificial.

Related Model Updates

The release notes document several concurrent changes to OpenAI's model lineup:

• GPT-5.4 mini rollout: Now available in ChatGPT as a fallback for GPT-5.4 Thinking during rate limits for Free and Go users, and automatically used for paid tiers when rate limits are reached.
• GPT-5.4 Thinking improvements: Enhanced reasoning capabilities, upfront thinking plans that allow mid-response adjustments, better web research for specific queries, and improved context window management.
• GPT-5.3 Instant earlier update (March 3): Delivered more accurate answers, richer web search results, and reduced unnecessary caveats.

Model Deprecations

OpenAI is retiring several legacy models to streamline its product lineup:

• GPT-5.1 models (March 11, 2026): GPT-5.1 Instant, Thinking, and Pro are no longer available. Existing conversations automatically migrate to GPT-5.3 Instant, GPT-5.4 Thinking, or GPT-5.4 Pro respectively.
• GPT-4o and older models (February 13, 2026): Retired from ChatGPT, including GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini. No API changes at this time.

Developer Actions

• ChatGPT users with conversations on GPT-5.1 will need to restart or switch models to access newer versions.
• GPT-5.4 mini will not appear as a selectable model in the UI; GPT-5 Thinking mini will be retired as selectable in 30 days.
• API customers should monitor deprecation timelines for any relevant model transitions.