ChatGPT Model Updates
- OpenAI released GPT-5.3 and GPT-5.4 in ChatGPT, improving spreadsheet work, frontend code, slides, maths, and document understanding. - ChatGPT can now search the web using an uploaded image, and creators can choose from the full model set in products. - Release notes also show several older model variants retired while API access stayed unchanged, affecting how platforms surface model choices ( ).
OpenAI has replaced ChatGPT’s older top-tier models with GPT-5.3 Instant and GPT-5.4 Thinking, shifting the app toward a two-track model picker and an automatic “Auto” mode. (help.openai.com) The company says GPT-5.3 Instant is now the default for logged-in users, while GPT-5.4 Thinking is the heavier-duty option for harder tasks. OpenAI says Instant improves info-seeking questions, how-tos, technical writing, and translation, and says Thinking is stronger on spreadsheets, frontend code, slides, maths, document understanding, image understanding, tool use, and web research. (help.openai.com) In ChatGPT Business, the model picker now centers on Auto, Instant, and Thinking, with Auto switching between Instant and Thinking on its own. OpenAI’s Business documentation says Instant has a 32K context window, while Thinking and Pro have 196K, and says GPT-5.4 Thinking and GPT-5.2 Thinking share a 3,000-requests-a-week cap, while Pro has a 15-requests-a-month cap. (help.openai.com) The update also changes how ChatGPT decides when to use extra reasoning. OpenAI says selecting Instant can still route a complex prompt to GPT-5.4 Thinking, and says users can add instructions while the model is still thinking before the final answer is complete. (help.openai.com) The model shuffle comes with a cleanup of older choices. OpenAI says ChatGPT retired GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and Thinking on February 13, 2026, then retired GPT-5.1 Instant, Thinking, and Pro on March 11, 2026. (help.openai.com) Those retirements apply to ChatGPT, not the application programming interface, or API, that outside developers use to build products. OpenAI says the retired models “will continue to be available through the OpenAI API,” while chats, projects, and GPTs inside ChatGPT are being moved to GPT-5.3 Instant or GPT-5.4 equivalents. (help.openai.com) That split leaves two different maps of OpenAI’s model lineup at the same time: a narrower set inside ChatGPT and a broader set for developers in the API. OpenAI’s GPT-building guide says retired ChatGPT models are no longer available there, even as the API help pages continue to describe “a range of model options” for builders. (help.openai.com, help.openai.com) OpenAI is also widening what ChatGPT can do with tools. Its capabilities guide says ChatGPT can analyze uploaded images, diagrams, screenshots, and charts, while its search guide says ChatGPT can automatically search the web when a question may benefit from current information. (help.openai.com, help.openai.com) The practical effect is less about one new model name than about how ChatGPT is being packaged: fewer legacy options, more automatic routing, and more tool use inside a single chat window. OpenAI’s own release notes now read less like a static menu and more like a moving system that decides, prompt by prompt, which model does the work. (help.openai.com, help.openai.com)
