Overview
- GPT-5.2 is now rolling out to paid ChatGPT plans and the API with a gradual deployment, while GPT-5.1 remains available under legacy options for three months before sunset.
- The lineup includes Instant for fast everyday tasks, Thinking for complex structured work such as coding, long‑document analysis and math, and Pro for maximum accuracy on difficult problems.
- OpenAI cites new highs on internal benchmark charts across coding, math, vision, long‑context reasoning and tool use; the company says Thinking reduces errors by 38% versus its predecessor and often tops Gemini 3 and Claude Opus 4.5 on reasoning tests.
- The release follows an internal ‘code red’ push to improve ChatGPT and counters Google’s Gemini 3 momentum, with OpenAI positioning GPT-5.2 as a default foundation for developers, agents and enterprise workflows.
- TechCrunch reports large infrastructure commitments and rising compute spend that raise scaling pressure, while OpenAI introduces safety steps around mental‑health use and teen age verification and signals a possible image‑focused model next month as an unconfirmed plan.