**GEPA**

GEPA (Genetic-Pareto) is an open-source framework for optimizing any system with textual parameters against any evaluation metric. Unlike RL or gradient-based methods that collapse execution traces into a single scalar reward, GEPA uses LLMs to read full execution traces—including error messages, profiling data, and reasoning logs—to diagnose why a candidate failed and propose targeted fixes. Through iterative reflection, mutation, and Pareto-aware selection, GEPA evolves high-performing variants with minimal evaluations. It has demonstrated 90x cost savings compared to proprietary models, 35x faster optimization than RL, and accuracy improvements from 32% to 89% on tasks such as prompt optimization (e.g., AIME benchmark), agent architecture discovery (ARC-AGI), cloud scheduling policy optimization, and coding agent skill learning. GEPA is used in production by companies including Shopify, Databricks, Dropbox, OpenAI, Pydantic, MLflow, and Comet ML.