OpenAI o3 and o4-mini: reasoning models learn to use tools
In one sentence OpenAI ships o3 (full) and o4-mini as reasoning models with native access to all ChatGPT tools: web search, Python, image gen, vision. First real 'agentic reasoning'.
OpenAI completes the transition from the o1 "preview" (September 2024) to real reasoning models. Two new ones:
- o3: the flagship for thinking long about hard problems (math, coding, science);
- o4-mini: the fast, cheap version, beyond o3-mini's capabilities.
The most important new thing: these models can use tools while thinking. For example, if you ask for an analysis of a PDF, o3 can: open the PDF, read it, plot charts with Python, search the web for updated info, then answer. All of this inside a single "thinking", not in separate calls.
For ChatGPT users, the "Reason" button becomes much more powerful. For developers, the API exposes tools as functions the model calls autonomously.
Companies
OpenAI
Tools
o3, o4-mini, o4-mini-high
Tags
Sources