GPT-5.2 (Non-reasoning) is a frontier large language model developed by OpenAI and released in December 2025. It is a configuration of the GPT-5.2 model series that operates without the extended internal chain-of-thought processing used by the "Thinking" and "Pro" variants. This version is optimized for low-latency performance and high-speed execution of professional knowledge tasks, such as spreadsheet modeling, code generation, and complex document analysis.
The model features a 400,000-token context window, allowing it to process and maintain continuity across extensive datasets or long-running conversations. It is inherently multimodal, supporting both text and image inputs while providing improved reliability in tool-calling and agentic workflows compared to its predecessor, GPT-5.1.
Benchmark evaluations indicate that GPT-5.2 (Non-reasoning) delivers significant gains in structured data handling and multi-step logic. It has been tested extensively on professional benchmarks like GDPval and GPQA Diamond, where it demonstrates high-level competency in domain-specific tasks spanning finance, legal, and scientific research. While it lacks the deeper iterative reasoning of its sibling models, it is engineered to minimize hallucinations and provide more consistent adherence to complex system instructions.