OpenAI Finally Drops GPT-5.4 — Watch AI Use Your Computer

OpenAI has officially launched GPT-5.4, rolling it out across ChatGPT, Codex, and its developer API. The company calls it its "most capable and efficient frontier model for professional work."

What's New in GPT-5.4?

Native Computer Use

For the first time, a general-purpose OpenAI model ships with native computer-use capabilities. GPT-5.4 can autonomously operate applications, issue mouse and keyboard commands in response to screenshots, and carry out multi-step workflows across software — without a human in the loop.

On OSWorld-Verified (desktop navigation benchmark), GPT-5.4 hit 75.0% success, up from 47.3% for GPT-5.2 — surpassing reported human performance of 72.4%.

Massive Context Window

The API version supports up to 1 million tokens of context, enabling agents to plan and execute long, complex tasks without losing track of earlier steps.

Fewer Errors & Hallucinations

• Individual responses are 33% less likely to contain errors vs. GPT-5.2
• Overall responses are 18% less likely to make mistakes
• Hallucinations are further reduced across benchmarks

Token Efficiency

GPT-5.4 solves equivalent tasks with 47% fewer tokens on some benchmarks, cutting costs and latency for developers.

Tool Search

A new API feature called Tool Search lets the model look up tool definitions only when needed — rather than loading all definitions upfront — reducing token waste in large tool ecosystems.

Who Gets Access?

Why It Matters

AI is rapidly moving from chatbot to agent. With native computer use, GPT-5.4 marks a meaningful shift: the model doesn't just answer questions — it can act on your behalf across your entire desktop environment. This has profound implications for automation, productivity, and how we think about AI in the workplace.