Benchmarks / Agentic
HANDBOOK.md Agents
Uses Surge AI's first-party MCP-based enterprise policy-following agent score. Score: score. Collected by Jake Boggs' benchmark dashboard.
HANDBOOK.md Agents
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Fable 5.1 | max | 38.8% |
| 2 | Fable 5 | max | 36.2% |
| 3 | Opus 5 | max | 32.3% |
| 4 | Grok 4.6 | xhigh | 27.3% |
| 5 | DeepSeek V4 Pro | max | 26.9% |
| 6 | GPT-5.6 Sol | max | 23.5% |
| 7 | DeepSeek V4 Flash | max | 23.1% |
| 8 | Opus 4.8 | max | 21.9% |
| 9 | GPT-5.5 | xhigh | 21.5% |
| 10 | Qwen 3.8 Max | xhigh | 16.5% |
| 11 | Grok 4.5 | high | 15.8% |
| 12 | Muse Spark 1.3 | xhigh | 15.4% |
| 13 | GLM 5.3 | max | 15.0% |
| 14 | Muse Spark 1.1 | xhigh | 13.5% |
| 15 | Muse Spark 1.2 | xhigh | 13.1% |
| 16 | Kimi K3 | max | 11.9% |
| 17 | Gemini 3.5 Flash | high | 11.2% |
| 18 | Gemini 3.7 Flash | high | 10.8% |
| 19 | Gemini 3.8 Flash | high | 10.8% |
| 20 | Sonnet 4.6 | max | 10.4% |
| 21 | GLM 5.3 Flash | max | 10.4% |
| 22 | Gemini 3.1 Pro | high | 10.0% |
| 23 | GLM 5.2 | xhigh | 10.0% |
| 24 | Qwen 3.7 Max | max | 8.5% |
| 25 | Hy3 | high | 7.7% |
| 26 | Kimi K2.6 | max | 6.9% |
| 27 | Gemini 3.6 Flash | high | 5.0% |
| 28 | Gemini 3.5 Flash-Lite | high | 3.1% |
| 29 | Inkling | xhigh | 2.3% |
| 30 | Grok 4.3 | high | 1.9% |
| 31 | Nemotron 3 Ultra 550B A55B | max | 1.5% |
| 32 | Inkling Small | xhigh | 1.2% |