OPENAI ANNOUNCES O3 AND O3-MINI, ITS NEXT SIMULATED REASONING MODELS
OpenAI's latest AI 'reasoning models', o3 and o3-mini, are now available for public safety testing and research access. The model family was called 'o3' rather than 'o2' to avoid potential trademark conflicts with British telecom provider O2. o3-mini will offer low, medium, and high processing speeds, with the higher compute settings producing better results. It outperforms o1 on the Codeforces benchmark.