ChatGPT's new model is out — GPT-5.4
OpenAI released GPT-5.4 today. It combines coding, reasoning, and computer use into one model for the first time. Here's what's new: - Computer use. GPT-5.4 is OpenAI's first general-purpose model with native computer-use capabilities. It can operate software, use a mouse and keyboard, and carry out tasks across applications through screenshots. - Thinking mode. GPT-5.4 Thinking shows you an upfront plan of what it's about to do, and you can adjust it mid-response while it's still working. - 1 million token context window (available in Codex, experimental). - Fewer errors. OpenAI says individual claims are 33% less likely to be wrong compared to GPT-5.2, and full responses are 18% less likely to contain mistakes overall. Benchmarks. It scored 75% on OSWorld (desktop navigation), which is above the human baseline of 72.4%. On professional knowledge tasks across 44 professions, it hit 83% vs 70.9% for GPT-5.2. Who gets it: GPT-5.4 Thinking is available for Plus, Team, and Pro users in ChatGPT. GPT-5.4 Pro is available through the API and for Enterprise and Edu subscribers. What's going away: GPT-5.2 Thinking moves to "Legacy Models" and will be discontinued on June 5, 2026. https://openai.com/index/introducing-gpt-5-4/ Adding this to my weekend list of things to test, let me know if you've tried it yet!