Earlier this week, OpenAI released GPT-5.3 Instant to make ChatGPT responses sound more natural. Now the company has introduced GPT-5.4 Thinking and GPT-5.4 Pro. The update improves its most advanced frontier model and focuses on coding, reasoning, and complex task automation.
OpenAI last updated this model in December with GPT-5.2, which arrived during rising competition from Google’s Gemini models. GPT-5.4 now pushes the model further with improvements across coding, document analysis, and agent workflows.
Improvements in GPT-5.4
GPT-5.4 improves several core capabilities. These include coding, instruction following, document understanding, and tool usage. The model also handles image perception and multimodal tasks more effectively.
It performs better in long-running workflows where an agent completes multiple steps. Teams can also run tool-heavy workloads with better token efficiency and faster overall performance.
The model also improves agentic web search and multi-source information synthesis. This helps users locate hard-to-find information across multiple sources. Businesses can also use GPT-5.4 for document-heavy workflows in customer service, analytics, and finance.
“GPT-5.4 is our most capable frontier model yet, delivering higher-quality outputs with fewer iterations across ChatGPT, the API, and Codex.”
GPT-5.4 supports a 1 million token context window. This allows users to analyze entire codebases, large document collections, or extended agent tasks in a single request.
OpenAI also added built-in computer use. This feature allows AI agents to interact directly with software and complete tasks through a build-run-verify-fix process.
“GPT-5.4 is the first mainline model with built-in computer-use capabilities, enabling agents to interact directly with software to complete, verify, and fix tasks.”
OpenAI said GPT-5.4 is rolling out gradually in ChatGPT and Codex. GPT-5.4 Thinking will replace GPT-5.2 Thinking for Plus, Team, and Pro users within three months.
