UniClawBench: leaderboard
Metric: Pass rate (%; tasks reaching a final pass state, mean of the English and Chinese subsets) on 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 |
|---|---|---|
| 1 | Claude Opus 4.8 | 47.5 |
| 2 | Claude Sonnet 4.6 | 45.5 |
| 3 | GPT-5.4 | 40.7 |
| 4 | Kimi K2.6 | 36.2 |
| 5 | Gemini 3.1 Pro (Preview) | 32.5 |
| 6 | Qwen 3.5 Plus | 31.8 |
| 7 | Gemini 3 Flash | 28.2 |
| 8 | GPT-5.4 Mini | 21.5 |
| 9 | Gemini 3.1 Flash Lite | 15.5 |
| 10 | GPT-4.1 | 15.2 |
Interactive version: theaggregate.ai/benchmark?slug=uniclawbench · How It Works · Data refreshed daily, snapshot 2026-09-29.