OpenAI introduces GPT-5.2, a new generation of AI models designed to make knowledge workers more productive.
With GPT-5.2, OpenAI is focusing on long-running agent tasks and claims that the new AI is “the most capable model series to date for professional knowledge work.”
Focused on knowledge work
OpenAI states that the model handles complex projects that require multiple steps better and is more reliable when using tools such as code environments, data sources, and document analysis.
Companies like Notion, Shopify, Databricks, and Zoom are already testing GPT-5.2 and report that the model is particularly good at long-term reasoning and handling tasks independently. The model is also said to perform better than its predecessor in programming work, such as recognizing bugs and adapting existing code.
Fewer errors, more context
According to OpenAI, GPT-5.2 hallucinates less and is better at combining information across long texts, such as contracts, reports, or research files. This would make the model more suitable for analysis, summarization, and decision-making, although human control remains necessary for critical work.
GPT-5.2 Instant, Thinking, and Pro are available in ChatGPT for paying users starting today. The models are immediately accessible to all developers via the API. GPT-5.1 can still be used for three months and will then be phased out.