OpenAI reasoning model that spends inference-time compute on step-by-step problem solving.
Specifications
- Provider
- OpenAI
- Type
- Vendor / proprietary
- Modality
- Reasoning
- Category
- Reasoning model
- Context window
- 200K
- License
- Proprietary (API)
- Released
- April 16, 2025
What it was trained for
A reasoning-focused model trained to think through problems step by step before answering, optimized for hard analytical and technical tasks.
Best for
- ▸Advanced math and scientific problem solving
- ▸Competitive-level coding and debugging
- ▸Multi-step logical analysis
- ▸Research and technical planning
- ▸Tasks requiring careful verification
Capabilities
Extended chain-of-thought reasoningVision inputFunction calling / toolsTool use during reasoning
Performance & positioning
Strong on demanding reasoning, math, and coding tasks; trades higher latency and cost for deeper deliberation than general chat models.
More from OpenAI