AuthorizationBench - OpenHands - Safe Completion Rate: leaderboard
AuthorizationBench's safe completion rate for OpenHands SDK 1.49.2 on 286 executable business tasks. Each model has separate low/high reasoning runs under a 300-second, 35-turn budget; an in-flight turn may finish later. Safe completion requires finishing without acting outside user authorization. All assigned trajectories remain in the denominator; MiMo v2.6 Pro supplies oracle-user replies and Claude Opus 5.5 judges outcomes. Source API/harness/effort identities are preserved.
Metric: Safe Completion Rate (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (High) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 83.9 | |
| 2 | GPT-5.6 Sol (Low) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 80.8 | |
| 3 | Claude Sonnet 5 (Low) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 79.7 | |
| 4 | Claude Sonnet 5 (High) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 77.3 | |
| 5 | Gemini 3.8 Flash (Low) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 74.8 | |
| 6 | GLM-5.3 (Low) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 73.8 | |
| 7 | GLM-5.3 (High) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 71.3 | |
| 8 | Gemini 3.8 Flash (High) (AuthorizationBench Run: OpenHands SDK 1.49.2; 300s; 35 turns) | 49 |
Interactive version: theaggregate.ai/benchmark?slug=authorizationbench-openhands-safe-completion-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.