We achieved gold medal level performance on this year's IMO! Our model thinks and writes proofs in clear, plain‑English - no formal code required. Unlike the narrower systems used in past competitions, our model is built to reason broadly, far beyond contest problems.

We competed live in the @atcoder World Finals Heuristic Contest and came in 2nd! Models like o3 rank among the top-100 in coding / math contests, but AFAIK this is the first top-3 placement in a premier coding / math contest. There's a huge gap between top-100 and top-3!

I am extremely excited about the potential of chain-of-thought faithfulness & interpretability. It has significantly influenced the design of our reasoning models, starting with o1-preview. As AI systems spend more compute working e.g. on long term research problems, it is critical that we have some way of monitoring their internal process. The wonderful property of hidden CoTs is that while they start off grounded in language we can interpret, the scalable optimization procedure is not adversarial to the observer's ability to verify the model's intent - unlike e.g. direct supervision with a reward model. The tension here is that if the CoTs were not hidden by default, and we view the process as part of the AI's output, there is a lot of incentive (and in some cases, necessity) to put supervision on it. I believe we can work towards the best of both worlds here - train our models to be great at explaining their internal reasoning, but at the same time still retain the ability to occasionally verify it. CoT faithfulness is part of a broader research direction, which is training for interpretability: setting objectives in a way that trains at least part of the system to remain honest & monitorable with scale. We are continuing to increase our investment in this research at OpenAI.

OpenAI Podcast Episode 2 is now live! @markchen90 and @nickaturley join @andrewmayne to pull back the curtain on the making of ChatGPT. They also get into how products are developed and what’s next for agentic coding and multimodal assistants.

Best reason to switch from using 4o to o3 is that o3 agrees the best running back in the NFL is @saquon!

Nothing beats burning the midnight oil in a maximum state of flow...

We don’t overhire on research. And on talent bar - there are some really phenomenal people joining over the next month! I spent years in trading until lack of meaning wore me down. I’ve never felt that at OAI - it’s so much more fulfilling to build things with real world impact.

Just released Codex, a software engineering agent that can work on many tasks in parallel. It runs on its own cloud-based compute infrastructure, and can fix bugs, answer questions about your code, run tests, etc.. Feels like a step towards the future of software engineering.

Product + Research = ❤️ Can’t wait to work with @markchen90 and bring his breakthroughs to life through our products.

So excited to work with @fidjissimo - she's a product visionary!