PawBench - Hermes: leaderboard
PawBench measures model capability under three fixed agent harnesses on AgentScope's task suite; this board reports overall score with the Hermes harness.
Metric: Overall Score (%). Source: agentscope-ai.github.io. Saturation forecast: Around January 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 78.44 | #63 |
| 2 | Qwen 3.7 Max | 72.29 | #75 |
| 3 | DeepSeek V4 Pro | 72.06 | #102 |
| 4 | Qwen 3.6 Plus | 70.41 | #120 |
| 5 | Qwen 3.6 27B | 68.2 | #170 |
| 6 | Qwen 3.6 Max Preview | 68.09 | #85 |
| 7 | Kimi K2.6 | 66.36 | #104 |
| 8 | GLM-5.1 | 63.19 | #97 |
| 9 | Qwen 3.6 35B A3B | 56.74 | #226 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pawbench-hermes · How It Works · Data refreshed daily, snapshot 2026-10-20.