o4-mini is the latest and fastest model in OpenAI's reasoning series, launched April 2025. It adds native image understanding to the reasoning pipeline — the model can look at a diagram, screenshot, or photo and reason over it just as it would over text. It sets new records on AIME 2025 and GPQA Diamond.
Key capabilities
- Reasoning + vision — can inspect and reason over images and diagrams
- Record-setting math — 99.5% on AIME 2025 competition problems
- Efficient — fast enough for interactive use while maintaining deep reasoning
- 200K context window with up to 100K output tokens
What it's best for
Coding agents, math tutoring, scientific Q&A that includes figures, and any pipeline needing strong reasoning with image comprehension. The go-to choice for Claude Code-style agentic coding loops requiring vision.