Overview

OpenAI has released GPT-5.4, its most capable and efficient frontier model to date, now available in ChatGPT (as GPT-5.4 Thinking and GPT-5.4 Pro), the API, and Codex. The model represents a significant advancement in reasoning, coding, and agentic workflows, combining industry-leading coding capabilities with improved performance across professional tasks like spreadsheet modeling, presentation creation, and document editing.

Key Capabilities

Knowledge Work & Professional Tasks

• Achieves 83.0% win rate on GDPval, matching or exceeding industry professionals across 44 different occupations
• 87.3% accuracy on investment banking spreadsheet modeling tasks (vs 68.4% for GPT-5.2)
• Human raters preferred GPT-5.4-generated presentations 68.0% of the time over GPT-5.2
• 33% less likely to make factual errors, and 18% less likely to contain any errors in full responses

Computer-Use & Agent Capabilities

• First general-purpose model with native, state-of-the-art computer-use capabilities
• Achieves 75.0% success rate on OSWorld-Verified (exceeding human performance at 72.4%)
• Supports up to 1M tokens of context, enabling long-horizon planning and execution
• Can operate computers via Playwright, mouse, and keyboard commands in response to screenshots
• New tool search feature helps agents find and use the right tools more efficiently

Performance & Efficiency

• Most token-efficient reasoning model yet; significantly fewer tokens required compared to GPT-5.2
• Better deep web research for highly specific queries
• Can now provide upfront reasoning plans in ChatGPT, allowing mid-response course corrections

Developer Actions

• ChatGPT users: Access GPT-5.4 Thinking or GPT-5.4 Pro through the interface; Enterprise customers can use the new ChatGPT for Excel add-in
• API developers: Integrate GPT-5.4 for computer-use agents and long-horizon tasks; updated spreadsheet and presentation skills available in the API
• Codex users: Access new computer-use capabilities and improved professional task handling