OpenAI released GPT-5.2. In ChatGPT it appears as instant and thinking modes plus a top-end GPT-5.2 Pro, while the API offers GPT-5.2 and GPT-5.2 Pro; a coding-focused GPT-5.2-Codex shipped alongside. The context window is 400,000 tokens with up to 128,000 tokens of output, and the knowledge cutoff moved to 31 August 2025, nearly a year on from GPT-5.1's 30 September 2024. API pricing is $1.75 per million input tokens and $14 per million output, 1.4 times GPT-5.1 ($21 and $168 for Pro). It scored 70.9% on GDPval, which measures realistic knowledge work, against 38.8% for GPT-5; 52.9% on the abstract-reasoning ARC-AGI-2 and 100% on the AIME 2025 competition maths set, against 17.6% and 94.0% for GPT-5.1 Thinking. Responses containing an error were, OpenAI said, 30% less common in relative terms than with GPT-5.1 Thinking. This was the model Altman's 1 December internal memo had referred to when it said a new reasoning model due the following week beat Gemini 3 in internal evaluations.