UniClawBench - Exploration: leaderboard
UniClawBench (HKU MMLab, 2026) runs proactive agents on 400 bilingual real-world tasks in live Docker containers under the OpenClaw framework, with up to two simulated user follow-ups and a hidden GPT-5.4 supervisor that grades step checkpoints. This board is the pass rate on the 80 exploration tasks, in percent; higher is better.
Metric: Pass rate (%) on the 80 exploration tasks of 400 bilingual real-world tasks run live in Docker containers under the OpenClaw framework (v2026.3.11), with up to two simulated user follow-ups and a hidden GPT-5.4 Codex supervisor that grades step checkpoints. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 85 | #94 |
| 2 | Claude Opus 4.8 | 82.5 | #56 |
| 3 | Claude Sonnet 4.6 | 81.2 | #161 |
| 4 | Gemini 3 Flash | 80 | #164 |
| 5 | Qwen 3.5 Plus | 78.7 | #232 |
| 6 | GPT-5.4 | 77.5 | #123 |
| 7 | Kimi K2.6 | 77.5 | #178 |
| 8 | GPT-5.4 Mini | 68.8 | #389 |
| 9 | Gemini 3.1 Flash Lite | 60 | #364 |
| 10 | GPT-4.1 | 51.2 | #495 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=uniclawbench-exploration · How It Works · Data refreshed daily, snapshot 2026-11-13.