LiveClawBench - Medium: leaderboard

Agent performance on the 46 Medium cases of LiveClawBench (2026), personal-assistant agent tasks in reproducible full-stack mock services; a case is Medium when the anchor model DeepSeek-V4-Pro earns a mean reward between 0.3 and 0.7 on it. Average reward over three runs, from 0 to 100; higher is better.

Metric: Average reward (0-100) over three runs on the 46 Medium cases (DeepSeek-V4-Pro anchor reward 0.3 to 0.7), 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. 16 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.567.7#54
2Kimi K2.7 Code67.3#203
3GLM-5.265#150
4GLM-5.163.2#198
5MiMo-V2.5-Pro62.6#284
6MiniMax-M359#280
7Qwen 3.6 Plus57.3#238
8Qwen 3.6 Flash57.2#448
9Claude Opus 4.856.6#65
10Kimi K2.655#192
11MiniMax-M2.748.9#539
12DeepSeek V4 Flash48.1#267
13Qwen 3.5 Plus37.6#242
14Qwen 3.5 Flash34.6#441
15Qwen 3.6 27B33.4#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-medium · How It Works · Data refreshed daily, snapshot 2026-11-21.