OpenAI introduced a new series of AI models, GPT-5.2, aimed at developers and professionals. The models are already available in ChatGPT for users with paid plans, and developers have access via API, where they can choose between “Instant,” “Thinking,” and “Pro” versions. The cost of using the API is $1.75 per million input tokens and $14 per million output tokens for the standard model, and for Pro — $21 and $168 respectively.
GPT-5.2 has received improvements in working with spreadsheets, creating presentations, programming, image processing, and understanding long texts. In ChatGPT, the model automatically switches between “Instant” and “Thinking” modes depending on the complexity of the task, and paid subscriptions allow working with context up to 196,000 tokens. The Pro version lacks support for “Canvas” features and image generation.
OpenAI notes a reduction in errors in GPT-5.2 compared to the previous version, improved resilience to harmful prompts, and updated behavior regarding user mental state signals. An age determination system for protecting minors has also been implemented, which will be gradually introduced.
Initial feedback from testers, including Windsurf and Triple Whale, indicates improvements in programming and tool usage. The previous version, GPT-5.1, will remain available in ChatGPT for another three months. GPT-5.2 models operate on Microsoft Azure infrastructure and NVIDIA GPUs.
