

Dspy · since Paper: arXiv 25. Juli 2025 (arXiv:2507.19457); DSPy 3.0 (August 2025) mit GEPA-Optimizer · 2× · last seen Nov 17, 2026
38
Momentum
GEPA (Genetic-Pareto) is a reflective prompt optimizer integrated into DSPy as dspy.GEPA and also available as a standalone Python library (pip install gepa). It analyzes execution traces of an LLM program, proposes instruction revisions in natural language, and selects candidates via Pareto-based selection. The source code is released under the MIT license.
Momentum trend
24.08.22.11.
Features
| License | MIT |
| Platform | Python (Python 3.10–3.14 for gepa[full]; DSPy installs gepa only on Python < 3.15) |
| Price | No price found (open-source library) |
| Protocol Compatibility | Not verified |
| Release Date | Paper arXiv:2507.19457 (2025); DSPy 3.4.1 on October 12, 2026 |
| Supported Models/Providers | Any LLM via dspy.LM; reflection LM freely selectable (e.g. openai/gpt-5); reflection LM must return JSON |