Overview

OpenAI introduced o1 as a preview in September 2024 and released the full model in December 2024. It was trained with reinforcement learning to reason through a hidden chain of thought before responding, and OpenAI reported large gains over GPT-4o on competition math, coding and PhD-level science questions. It launched a new generation of “reasoning models” across the industry.

Highlights

  • Extended internal reasoning before answering
  • Strong results on math, coding and science benchmarks at release
  • Image input in the full release
  • Large context window in the API
  • Succeeded by newer o-series and GPT reasoning models

Typical uses

Complex problem solving in math, science, programming and planning, where accuracy matters more than speed.

Things to consider

  • Slower and more expensive per answer than standard chat models.
  • The internal reasoning is not fully visible to users.
  • Newer OpenAI models have superseded it for most uses.

Specifications are taken from the provider’s official announcement, documentation or model card. Benchmark results depend on evaluation settings; see how AI benchmarks work.