Benchmarks · agentic
HiL-Bench
Uses Scale Labs combined Pass@3 with ask_human tool. Score: combined Pass@3 with ask_human tool. Collected by Jake Boggs' benchmark dashboard.
31 models · 31 results · data as of 2026-11-06 · labs.scale.com ↗
Leaderboard
Muse Wonder (Max) max
All results
31 rows| # | Model | Effort | HiL-Bench |
|---|---|---|---|
| 1 | Muse Wonder (Max) | max | 79.7% |
| 2 | GPT-6.1 Terra | max | 68.0% |
| 3 | Muse Spark 1.3 | max | 66.9% |
| 4 | GPT-6 Astra | max | 65.3% |
| 5 | GPT-6.1 Sol | max | 62.3% |
| 6 | Gemini 4 Argon | max | 62.0% |
| 7 | Fable 5.1 | max | 61.5% |
| 8 | SWE 2 | max | 61.5% |
| 9 | Muse Spark Open | xhigh | 59.7% |
| 10 | Opus 5.5 | max | 59.0% |
| 11 | GLM 5.3 | max | 58.3% |
| 12 | Opus 5 | max | 57.0% |
| 13 | Fable 5 | max | 56.3% |
| 14 | DeepSeek V4 Flash | max | 47.8% |
| 15 | GLM 5.2 | max | 43.7% |
| 16 | Kimi K3 | max | 42.7% |
| 17 | Opus 4.7 | max | 41.7% |
| 18 | Gemini 3.8 Flash | max | 41.5% |
| 19 | GPT-5.5 | xhigh | 39.7% |
| 20 | DeepSeek V4 Pro | max | 38.5% |
| 21 | Opus 4.6 | max | 38.3% |
| 22 | MiniMax M3 | max | 38.3% |
| 23 | Opus 4.8 | max | 35.3% |
| 24 | Gemini 3.1 Pro | high | 35.3% |
| 25 | GPT-5.6 Sol | max | 32.3% |
| 26 | Gemini 3.5 Flash | xhigh | 27.7% |
| 27 | Grok 4.20 | xhigh | 20.0% |
| 28 | Kimi K2.6 | max | 18.7% |
| 29 | GPT-5.4 | xhigh | 9.7% |
| 30 | MiniMax M2.5 | high | 6.3% |