Paper • 2602.22953 • Published • 12
React
This is a tracking repo for React, used by the Open Agent Leaderboard to report evaluation results on HuggingFace.
Standard ReAct (Reason + Act) agent loop via LiteLLM with tool calling across any supported model.
- Framework: litellm
- Leaderboard: Open Agent Leaderboard
- Paper: arXiv:2602.22953
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Paper for Exgentic/react
Evaluation results
- open-agent-leaderboard/results
- Overall
- model: Claude Opus 4.5 View evaluation results source 0.61 *
- model: DeepSeek V3.2 View evaluation results source 0.46 *
- model: GPT-5.2 View evaluation results source 0.41 *