OpenAI o3 at a glance
o3 and o4-mini were trained to decide when to search, run code or inspect images during reasoning, replaced o1 and o3-mini in ChatGPT, and arrived with the open-source Codex CLI.
What changed with OpenAI o3
On April 16, 2025, OpenAI released o3, its most powerful reasoning model at the time, and o4-mini, a smaller model for fast, cost-efficient reasoning. OpenAI said o3 set new highs on Codeforces, SWE-bench and MMMU, and that o4-mini was the best benchmarked model on AIME 2024 and 2025.
Both models were trained with reinforcement learning to reason about when and how to use tools, with full tool access in ChatGPT and custom tools via function calling in the API. They can also bring images into their chain of thought, cropping or rotating a photo of a whiteboard or diagram while working. In ChatGPT they replaced o1 and o3-mini for paid plans, and free users could try o4-mini. OpenAI also open-sourced Codex CLI, a terminal coding agent.
Where to compare OpenAI o3
Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.
Leaderboards
General leaderboards
Why OpenAI o3 matters for businesses
For businesses, o3 and o4-mini marked the shift from chatbots to agents that look things up, analyze files and check their own work in one request. That makes them useful for research, data analysis and visual inspections, though longer reasoning adds latency and cost, so match the model to the task.
Sources for this story
- Introducing OpenAI o3 and o4-mini (opens in a new tab)Official
OpenAI · openai.com
- openai/codex (opens in a new tab)Official
GitHub · github.com
- o3 model page (opens in a new tab)Benchmark
Artificial Analysis · artificialanalysis.ai
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #llm
- #openai
- #reasoning
- #o-series
- #tool-use
- #coding
