Particle.news
Download on the App Store

OpenAI Launches GPT-5.4 With Native Computer Use and 1M-Token Context

OpenAI positions the update for professional, agentic workflows built on native computer control with a one‑million‑token context window.

Overview

  • GPT-5.4 is rolling out across ChatGPT, the API, and Codex, with GPT-5.4 Thinking for Plus, Team, and Pro users and a higher-performance GPT-5.4 Pro for Pro and Enterprise plans.
  • The system combines advanced reasoning and coding with built-in computer use, issuing mouse and keyboard commands from screenshots and outlining a plan that users can adjust mid-response.
  • OpenAI highlights stronger web research and multi-source synthesis, more accurate tool and API calling, and a new tool search that loads detailed tool definitions only when needed to cut token use.
  • The company says GPT-5.4 is its most factual model to date, with individual claims 33% less likely to be false and full responses 18% less likely to contain errors versus GPT-5.2, alongside token-efficiency gains.
  • Earlier this week, GPT-5.3 Instant became ChatGPT’s default quick-reply model with fewer unnecessary refusals and improved web contextualization, though OpenAI’s system card notes safety-benchmark regressions on some disallowed-content tests; GPT-5.2 remains accessible in legacy menus until June 3, 2026.