The Story

OpenAI has officially introduced GPT-5.4, describing it as its most capable and efficient frontier model for professional work. The company is rolling it out across ChatGPT as GPT-5.4 Thinking, the API as gpt-5.4, and Codex, while also offering GPT-5.4 Pro for users and developers who want maximum performance on complex tasks.

What is new in GPT-5.4

• Professional work focus: OpenAI says GPT-5.4 combines recent advances in reasoning, coding, and agentic workflows into one mainline model designed for real-world tasks.
• Computer use: The model is OpenAI’s first general-purpose release with native computer-use capabilities, allowing agents to operate across software and desktop-style environments.
• Long context: GPT-5.4 supports up to 1 million tokens of context in Codex and the API for long-horizon planning, execution, and verification.
• Tool efficiency: OpenAI says the model works more effectively across larger tool ecosystems while using fewer tokens than GPT-5.2 on many tasks.

Performance highlights

• Knowledge work: On GDPval, GPT-5.4 matched or exceeded professionals in 83.0% of comparisons across 44 occupations.
• Computer use: On OSWorld-Verified, GPT-5.4 reached a 75.0% success rate, ahead of GPT-5.2’s 47.3% and above the reported human baseline of 72.4%.
• Coding: On SWE-Bench Pro (Public), GPT-5.4 scored 57.7%, slightly ahead of GPT-5.3-Codex and GPT-5.2.

Why it matters

This launch shows OpenAI pushing beyond chatbot quality into broader workplace execution. Rather than positioning GPT-5.4 only as a smarter model, OpenAI is framing it as a system for handling spreadsheets, presentations, documents, coding workflows, and software interaction with less back-and-forth. That makes the release especially relevant for enterprises, developers, and teams building long-running AI agents.

Availability

GPT-5.4 Thinking is rolling out in ChatGPT for Plus, Team, and Pro users, replacing GPT-5.2 Thinking. GPT-5.4 is also available now in the API and Codex, with GPT-5.4 Pro offered for users who need maximum performance on complex tasks.