Krux
March 20, 2026
OpenAI's GPT-5.4 Brings 1M-Token Context to API
Published: March 20, 2026 at 12:58 AM
Updated: March 20, 2026 at 12:58 AM
100-word summary
OpenAI released GPT-5.4 in "Thinking" and "Pro" variants alongside GPT-5.3 Instant as the default fast option across ChatGPT and its API. The biggest change: developers get 1 million tokens of context in the API and Codex, enough to hold roughly 750,000 words or hundreds of pages of documentation in a single conversation. The update adds native computer-use capabilities and configurable reasoning modes. Pro-tier users get 400K-token windows when using GPT-5.4 Thinking, while free users cap at 16K. What you can do today that you couldn't yesterday: run long automation tasks that plan, execute, and verify across hours instead of minutes without losing the thread.
What happened
OpenAI released GPT-5.4 in "Thinking" and "Pro" variants alongside GPT-5.3 Instant as the default fast option across ChatGPT and its API. The biggest change: developers get 1 million tokens of context in the API and Codex, enough to hold roughly 750,000 words or hundreds of pages of documentation in a single conversation.
Why it matters
The update adds native computer-use capabilities and configurable reasoning modes. Pro-tier users get 400K-token windows when using GPT-5.4 Thinking, while free users cap at 16K. What you can do today that you couldn't yesterday: run long automation tasks that plan, execute, and verify across hours instead of minutes without losing the thread.