OpenAI introduced the o1 reasoning model. Unlike previous GPT models, o1 generates an internal "chain of thought" before responding, delivering significantly better results on complex reasoning, science, and math tasks.
On the 2024 AIME math exams, o1 scored 74% compared to GPT-4o's 12%. This represents a massive leap in reasoning capabilities.
OpenAI also launched a $200/month ChatGPT Pro subscription tier alongside the full o1 release, designed for professional users and researchers.