meituan-longcat/

AMO-Bench

📄 Paper 🌐 Project Page 💻 Github Repo

2026.02.05: Leaderboard Update: Qwen3-Max-Thinking achieves a new SOTA with 65.1%, while GLM-4.7 sets a new open-source record at 62.4%! 2025.12.01: We have added Token Efficiency showing the number of output tokens used by models in the leaderboard. Gemini 3 Pro achieves the highest token efficiency among top-performance models!… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/AMO-Bench.