o3 and o4-mini (OpenAI)
OpenAI launches o3 and o4-mini, reasoning models able for the first time to use tools within ChatGPT (web browsing, Python code, image analysis and generation) as they think.
Key points
- OpenAI launches o3 and o4-mini, reasoning models.
- For the first time, they use tools in ChatGPT during their thinking.
- These tools include web browsing, Python code and image analysis or generation.
- Assistants now combine reasoning and active web search.
Analysis
With o3 and o4-mini, ChatGPT can mobilize tools during its very reasoning, notably fetching information on the web when the thinking requires it. Search is no longer an isolated step but an action integrated into the answer process, which changes how sources are called upon.
For AI visibility, this reinforces the stake of being findable and citable at the moment the assistant reasons and decides to verify a point. Up-to-date, precise and easy-to-extract content has more chance of being consulted during reasoning and reused in the final answer.
What to do
- Ensure the accessibility and indexability of key pages so they can be consulted during active browsing.
- Maintain the freshness of factual content likely to be verified during an answer.
- Test complex queries in ChatGPT and observe whether your site is consulted and cited during reasoning.
Assistants combine reasoning and active web search. The ability to be found and cited during reasoning becomes key for AI visibility.