← Back
OpenAI
OpenAI releases GPT-5.4, advancing reasoning and agent capabilities with computer-use features
OpenAI APIChatGPTOpenAI · releasemodelfeatureapiplatform · openai.com ↗

Overview

OpenAI has announced GPT-5.4, their most capable frontier model to date, now available across ChatGPT (as GPT-5.4 Thinking and Pro), the OpenAI API, and Codex. The release brings together advances in reasoning, coding, and agentic workflows, designed specifically for professional work and complex task automation.

Key Capabilities

Reasoning & Interactivity: In ChatGPT, GPT-5.4 Thinking provides upfront visibility into its reasoning process, allowing users to adjust course mid-response before reaching a final output. The model also improves deep web research for highly specific queries and better maintains context for extended thinking tasks.

Computer Use: GPT-5.4 is the first general-purpose model OpenAI has released with native computer-use capabilities, enabling agents to operate computers and execute complex workflows across applications. It can write code using libraries like Playwright and issue mouse/keyboard commands in response to screenshots. Developers can configure safety behavior and confirmation policies to suit different risk tolerances.

Efficiency & Scale: The model supports up to 1M tokens of context for long-horizon task planning and execution. GPT-5.4 is OpenAI's most token-efficient reasoning model yet, using significantly fewer tokens than GPT-5.2 to solve problems, resulting in reduced costs and faster speeds. It also introduces tool search, helping agents discover and use the right tools more efficiently.

Performance Benchmarks

GPT-5.4 demonstrates substantial improvements across professional and technical benchmarks:

  • GDPval (knowledge work across 44 occupations): 83.0% win/tie rate vs. industry professionals (up from 70.9% for GPT-5.2)
  • SWE-Bench Pro: 57.7% (vs. 55.6% for GPT-5.2)
  • OSWorld-Verified (computer-use tasks): 75.0% (vs. 47.3% for GPT-5.2)
  • Toolathlon: 54.6% (vs. 46.3% for GPT-5.2)
  • Spreadsheet modeling: 87.3% mean score on junior analyst tasks (vs. 68.4% for GPT-5.2)
  • Factuality: Individual claims are 33% less likely to be false; full responses are 18% less likely to contain any errors

Professional Knowledge Work

GPT-5.4 shows significant improvements in creating and editing professional documents. Human raters preferred GPT-5.4-generated presentations over GPT-5.2 68% of the time, citing stronger aesthetics and visual variety. The model excels at spreadsheet modeling, presentation creation, and document handling tasks that professionals rely on daily.

Enterprise customers can now leverage the newly released ChatGPT for Excel add-in to integrate GPT-5.4's capabilities directly into Excel workflows. Updated spreadsheet and presentation skills are also available in Codex and the API.

Availability

GPT-5.4 is available immediately in ChatGPT and via the OpenAI API. A premium tier, GPT-5.4 Pro, offers maximum performance for complex tasks. Enterprise customers have access to the new Excel integration and updated professional skills modules.