Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
JSONSchemaBench is a benchmark of approximately 10,000 real-world JSON schemas designed to evaluate the efficiency and coverage of structured output generation from large language models. It measures how well constrained decoding engines or LLMs can produce outputs that conform to complex JSON Schema constraints.
Parse Score