GPT-5 nano (high) is a specialized configuration of the GPT-5 nano model, the most compact and efficient variant in OpenAI's GPT-5 series released in August 2025. This specific version is characterized by its use of the "high" reasoning effort setting, which allows the model to allocate more compute to internal processing for complex tasks compared to the default "medium" or speed-optimized "minimal" settings.
Despite its smaller size, the model maintains a massive 400,000-token context window, enabling the processing of extensive documents, codebases, and long-form conversations. It is designed to bridge the gap between low-latency performance and high-quality reasoning, making it suitable for real-time applications that require more than basic classification or summarization.
Architecturally, the model is part of OpenAI's unified system that utilizes intelligent routing to handle varying task complexities. While the flagship GPT-5 and GPT-5 Pro models handle extreme-scale reasoning, the nano variant with high effort is optimized for developer tools and agentic workflows where response speed remains a priority but logic cannot be compromised. Benchmarks indicate that it performs significantly above average for its size class in intelligence and verbosity.