Open Agent Leaderboard

An open benchmark for comparing full AI agent systems across diverse real-world tasks. Reports both quality and cost.

Unlike model-only benchmarks, we evaluate the complete agent — the model, the tools, the planning strategy, the error recovery — as a single system. The same model can produce very different results depending on the agent wrapped around it.

• Website: exgentic.ai
• Results: open-agent-leaderboard/results
• Leaderboard: open-agent-leaderboard/leaderboard
• Blog: open-agent-leaderboard/blog
• Framework: Exgentic
• Paper: arXiv:2602.22953

Submit results

Run evaluations using Exgentic and open a PR on the results dataset.