Claw-Eval-Live - Overall Completion: leaderboard

Workflow-agent task completion on Claw-Eval-Live (2026): 105 executable tasks built from public ClawHub workflow demand, 87 service-backed business workflows (CRM, finance, email, helpdesk) and 18 local workspace repairs, graded from execution traces, service audit logs and workspace state. Mean completion score from 0 to 100, which credits partial progress unlike the pass rate; higher is better.

Metric: Overall completion score (0-100): mean task score over the 105 released Claw-Eval-Live workflow tasks (87 service-backed business workflows, 18 local workspace repairs), each task scored from 0 to 1 by deterministic checks on traces, audit logs, service state and artifacts plus a GPT-5.4 rubric judge for semantic dimensions; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 4.683.6#114
2GPT-5.481.7#138
3Claude Sonnet 4.679.9#179
4GLM-578.1#276
5MiniMax-M2.777.5#539
6MiMo-V2-Pro76.9#356
7Kimi K2.576.2#272
8Gemini 3.1 Pro (Preview)74#102
9Qwen 3.5 397B A17B72.7#270
10Qwen 3.6 Plus71.4#245
11MiniMax-M2.570.9#600
12Doubao Seed 2.070.4
13DeepSeek V3.269.3#425

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=claw-eval-live-overall-completion · How It Works · Data refreshed daily, snapshot 2026-11-22.