OpenAI has launched GPT-5.4 Thinking, a significant upgrade to its flagship model that brings six key improvements across coding, document analysis, and multimodal tasks. The update follows GPT-5.3 Instant's release earlier this week and comes after GPT-5.2's December launch, which was OpenAI's response to Google's Gemini competition.

GPT-5.4 introduces a 1 million token context window, enabling users to analyze entire codebases or lengthy document collections in a single request. The model also debuts built-in computer use capabilities, allowing agents to interact directly with software through a build-run-verify-fix loop. OpenAI describes it as the first mainline model trained to support compaction for longer agent trajectories while preserving key context.

The upgrade is rolling out gradually in ChatGPT and Codex, with GPT-5.4 Thinking available to Plus, Team, and Pro subscribers. The previous GPT-5.2 Thinking model will be discontinued in three months. OpenAI emphasizes improvements in tool-heavy workloads, agentic web search, and document-heavy business workflows for customer service, analytics, and finance applications.