LiveClawBench - Pass@3: leaderboard

Agent reliability on LiveClawBench (2026), personal-assistant agent tasks in reproducible full-stack mock services (stateful web applications, browser, APIs and shell) with implicit goals and runtime perturbations, three runs per case scored by case verifiers and a rubric judge. Share of the 134 cases in which at least one of the three runs earns a reward above 0.8, in percent; higher is better.

Metric: Pass@3 (x100): share of the 134 cases where at least one of three runs has reward above 0.8, OpenClaw-style personal-assistant tasks run in reproducible full-stack mock services (22 stateful web applications, browser, APIs and shell), three runs per case, rewards from case verifiers and a DeepSeek-V3.2 rubric judge for open-ended cases; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 17 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Kimi K2.7 Code70.9#203
2GLM-5.169.4#198
3MiniMax-M368.7#280
4GPT-5.567.9#54
5DeepSeek V4 Pro67.9#196
6MiniMax-M2.766.4#539
7GLM-5.266.4#150
8Kimi K2.665.4#192
9Qwen 3.6 Flash64.2#448
10Qwen 3.6 Plus63.4#238
11MiMo-V2.5-Pro63.4#284
12Claude Opus 4.861.9#65
13DeepSeek V4 Flash61.9#267
14Qwen 3.5 Plus59.7#242
15Qwen 3.6 27B58.2#351

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=liveclawbench-pass-3 · How It Works · Data refreshed daily, snapshot 2026-11-21.