HiL-Bench: leaderboard
Scale Human-in-Loop benchmark measuring when agents should ask for help, escalate uncertainty, or continue autonomously.
Metric: Combined Pass@3 (%). Source: labs.scale.com. Saturation forecast: Around February 2027. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Wonder | 79.67 | #15 |
| 2 | GPT-6 Astra | 65.32 | #6 |
| 3 | GPT-6.1 Sol | 62.33 | #7 |
| 4 | Gemini 4 Argon | 62 | #8 |
| 5 | Claude Fable 5.1 | 61.5 | #9 |
| 6 | Claude Opus 5.5 | 59 | #3 |
| 7 | GLM-5.3 | 58.33 | #39 |
| 8 | Claude Opus 5 | 57 | #12 |
| 9 | Claude Fable 5 | 56.33 | #16 |
| 10 | DeepSeek V4 Flash | 47.81 | #137 |
| 11 | GLM-5.2 | 43.67 | #73 |
| 12 | Kimi K3 | 42.67 | #30 |
| 13 | Claude Opus 4.7 | 41.67 | #49 |
| 14 | Gemini 3.8 Flash | 41.47 | #21 |
| 15 | GPT-5.5 | 39.67 | #31 |
No result here: #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=hil-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.