LLM Stats (SkillsBench): leaderboard
LLM Stats mirror of SkillsBench, evaluating agentic and coding task performance when models use structured procedural skill packages across diverse domains.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 Max | 70.2 | #35 |
| 2 | Hy4 preview | 62.9 | #38 |
| 3 | Qwen 3.7 Max | 59.2 | #75 |
| 4 | Hy3 | 55.3 | #93 |
| 5 | Qwen 3.7 Plus | 54.9 | #92 |
| 6 | Qwen 3.6 27B | 48.2 | #170 |
| 7 | Qwen 3.6 Plus | 45.7 | #120 |
| 8 | Ling-3.0-flash | 44.8 | #206 |
| 9 | Muse Glimmer 30B | 44.3 | #178 |
| 10 | Qwen 3.6 35B A3B | 28.7 | #226 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-skillsbench · How It Works · Data refreshed daily, snapshot 2026-10-20.