Benchmarks / Agentic
SkillsBench
How important are skills for agents? Vals AI's overall SkillsBench score.
SkillsBench
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | DeepSeek V4.1 Flash | high | 69.8% |
| 2 | Grok 4.5 | high | 66.0% |
| 3 | Gemini 3.7 Flash | high | 65.9% |
| 4 | GPT-5.5 | xhigh | 62.6% |
| 5 | Fable 5.1 | max | 61.6% |
| 6 | GPT-5.6 Luna | max | 60.4% |
| 7 | Opus 5 | max | 60.4% |
| 8 | Opus 4.8 | max | 59.2% |
| 9 | Muse Spark 1.1 | xhigh | 59.2% |
| 10 | GPT-5.6 Terra | max | 58.9% |
| 11 | Gemini 3.8 Flash | high | 58.0% |
| 12 | Grok 4.6 | high | 55.8% |
| 13 | Qwen 3.7 Plus | high | 54.3% |
| 14 | GPT-5.6 Sol | max | 54.1% |
| 15 | DeepSeek V4 Pro 0813 | max | 53.8% |
| 16 | Muse Spark 1.2 | xhigh | 53.0% |
| 17 | Gemini 3.5 Flash | high | 52.7% |
| 18 | GPT-5.4 | xhigh | 51.7% |
| 19 | MiniMax M3 | max | 51.5% |
| 20 | DeepSeek V4 | max | 51.3% |
| 21 | DeepSeek V4 Flash 0731 | high | 50.7% |
| 22 | Kimi K2.7 Code | max | 50.0% |
| 23 | Sonnet 4.6 | max | 49.1% |
| 24 | GLM 5.3 | max | 47.5% |
| 25 | Sonnet 5 | max | 46.5% |
| 26 | GLM 5.2 | max | 45.1% |
| 27 | Qwen 3.8 Max | max | 42.0% |
| 28 | Grok 4.3 | high | 40.6% |
| 29 | GLM 5.3 Flash | max | 40.2% |
| 30 | Qwen 3.8 27B | xhigh | 38.1% |
| 31 | Inkling Small | xhigh | 33.6% |
| 32 | Ling 3.0 Flash | – | 27.9% |
| 33 | Inkling | max | 26.1% |
| 34 | Mercury 2.5 | high | 18.1% |