GPT-5.2 (medium) is a configuration of OpenAI's GPT-5.2 model series released in December 2025. It specifically refers to the model when operated with a medium reasoning effort setting, designed to provide a balance between inference speed and complex problem-solving capabilities. The model was launched as part of a series that includes Instant, Thinking, and Pro variants, following a rapid development cycle to address competition in multimodal and agentic performance.
Built for professional knowledge work and autonomous agentic workflows, the model features a significant 400,000-token context window and a 128,000-token maximum output capacity. This architecture allows for the processing of large codebases, multi-document analysis, and the generation of extensive technical documentation in a single pass. It utilizes a knowledge cutoff of August 31, 2025.
In terms of performance, the medium reasoning configuration is optimized for tasks such as spreadsheet creation, financial modeling, and multi-step project execution. It shows improved reliability and reduced hallucinations compared to earlier iterations, particularly in structured coding and analytical queries. The model series supports multimodal inputs and outputs, including text and image processing.