o3 trades speed for deliberation, which suits territory analysis and deal post-mortems far better than it suits drafting email.
Reasoning models are slower and cost more per answer, so the honest use is the small number of high-stakes questions — why did this quarter's forecast miss — not the daily drafting work.
OpenAI released o3 and o4-mini on 16 April 2025, the latest in its o-series of models trained to think for longer before responding. The release added reasoning over images and more capable agentic tool use, and introduced Codex CLI alongside.
OpenAI shipped o3-pro to Pro users on 10 June 2025, recorded as an update on the same announcement page. Both models can call tools during reasoning rather than only before or after it.
Two minutes, once a week. What changed in AI, and what to run because of it.