OpenAI launches GPT-5.4 with 1 million-token context window
OpenAI just rolled out GPT-5.4—their latest AI model—in March 2026. It comes in three flavors: Standard, Thinking (for deep reasoning), and Pro (for max performance).
If you're on ChatGPT Plus, Team, or Pro—or use the API—only the API supports a massive 1 million-token context window.
Tool search helps it grab tools when needed
GPT-5.4 crushes benchmarks with an 83% score on knowledge tasks and cuts down claim errors by a third compared to GPT-5.2.
It introduces "Tool Search," letting it grab tools when needed—so it's more efficient and saves on API costs.
Agents can now natively use your computer
While tokens cost a bit more than before, GPT-5.4's speed and accuracy help balance things out.
For the first time, agents can natively use your computer—browsing desktops or apps without extra setup.
Plus, new integrations with Excel, Google Sheets, and financial data tools make it handy for anyone juggling complex projects or data-heavy work.