o3 is an OpenAI reasoning model from April 2025 that thinks for longer before it answers and can decide when to use tools.
In plain words
o3 belongs to OpenAI's o-series, trained to work through a problem before replying. OpenAI scaled up reinforcement learning to train it and found that letting it think longer kept improving results. It was also trained to choose tools itself, such as web search or running Python, and it could reason about images as part of its thinking. Its dated API version is now marked deprecated.
A quick example
Upload a blurry photo of a whiteboard full of maths and ask for the answer. o3 can zoom into and study the image while reasoning, then work through the problem before writing its reply.
Where you'll meet it
- ChatGPT and the OpenAI API, from April 2025.
- Maths, science, coding and visual reasoning tasks.
Related
Sources
- DocsIntroducing OpenAI o3 and o4-mini, OpenAI · read 29 Sept 2026
- Docso3 model, OpenAI API docs · read 26 Oct 2026
Checked by WhatIsAI against the sources above · 26 Oct 2026. Spotted an error? Report it.