GPT-5.4

GPT-5.4 incorporates and improves upon the industry-leading coding capabilities of GPT-5.3-Codex, making it highly proficient in code generation, understanding, and debugging.

This is the first general-purpose model released with native, state-of-the-art computer-use capabilities, enabling AI agents to operate computers and execute complex workflows across various applications, including writing code for automation (e.g., Playwright) and issuing mouse/keyboard commands.

GPT-5.4 supports up to 1 million tokens of context, allowing agents to plan, execute, and verify tasks across significantly longer horizons and maintain extensive conversational or operational context.

Available in ChatGPT, GPT-5.4 Thinking provides an upfront plan of its thought process, allowing users to adjust the model's course mid-response. This feature aims to deliver more aligned final outputs with less back-and-forth interaction.

The model delivers improved and more consistent results on professional tasks, particularly in creating and editing spreadsheets, presentations, and documents. It achieves state-of-the-art performance on benchmarks like GDPval and internal spreadsheet modeling tasks.

GPT-5.4 is OpenAI's most factual model to date, demonstrating a 33% reduction in false individual claims and an 18% reduction in full responses containing any errors compared to GPT-5.2.

The model is designed to be highly token efficient, using significantly fewer tokens to solve problems compared to GPT-5.2, which translates to reduced token usage and faster processing speeds.