ChemGraph Leaderboard
An interactive benchmark of single- and multi-agent AI for computational chemistry.
ChemGraph Leaderboard provides a reproducible comparison of agentic AI frameworks and large language models for computational chemistry and materials science. It evaluates single- and multi-agent workflows on 40 ground-truth queries across 12 task categories, reporting accuracy alongside token usage and cost (Pham et al., 2026).
- Stack: Python, Gradio, Hugging Face Hub, ChemGraph.
- Live: huggingface.co/spaces/Autonomous-Scientific-Agents/chemgraph-leaderboard
- Code: github.com/Autonomous-Scientific-Agents/chemgraph-leaderboard