OpenAI has released GPT-5.4, its most capable and efficient model yet for professional work.
The new model combines advances in reasoning, coding, and agentic workflows building on the coding strengths of GPT-5.3-Codex while significantly improving tasks across spreadsheets, documents, presentations, and complex workflows.
Key highlights:
• State-of-the-art performance across knowledge work benchmarks (83% win/tie vs industry professionals)
• Native computer-use capabilities, enabling agents to operate software and complete multi-step tasks
• 1M token context window for long-horizon planning and workflows
• Stronger coding, faster performance, and lower token usage
• Improved factual accuracy with fewer hallucinations
GPT-5.4 is now rolling out across ChatGPT, the API, and Codex, with GPT-5.4 Pro available for maximum performance on complex tasks.
A big step toward more reliable AI agents and real-world productivity.
