OpenAI has released an update to GPT-5.1

The GPT-5.1 Instant model is optimized for quick responses and is suitable for everyday tasks: short conversations, email composition, and simple code requests. Unlike previous versions, it now automatically uses concise reasoning for complex questions. The response structure and communication style have also been improved.

The GPT-5.1 Thinking model is designed for complex tasks requiring deep analysis, such as multi-step computation, programming, and planning. It flexibly allocates computing resources based on the complexity of the query. Responses are now less jargon-intensive, more consistent, and retain context, although the performance is slower than Instant.

At the same time, the ChatGPT personalization system has been updated. Six communication styles have been added: Standard, Friendly, Effective, Professional, Direct, and Creative. The selected style applies to all chats. Fine-tuning of these settings is planned for the future.

Developers will retain backward compatibility with the APIs of previous GPT-5 versions. The new models have passed standard security tests and feature improved filters for emotionally sensitive scenarios. Baseline GPT-5 models will remain available for three months for comparison with the new versions.