ClawMark - Task Success: leaderboard
ClawMark (2026) runs coworker agents through 100 multi-day tasks in 13 professional scenarios on stateful sandboxed email, calendar, file, knowledge-base and spreadsheet services whose state changes between turns, scored by deterministic Python checkers; every model runs in the OpenClaw agent framework. This board is the share of tasks on which every checker passes, in percent; higher is better.
Metric: Task success rate (%): tasks on which every checker passes; 100 multi-day coworker tasks in 13 professional scenarios on five stateful sandboxed services (filesystem, email, calendar, knowledge base, spreadsheet) with multimodal evidence; every model runs in the OpenClaw agent framework, one sweep, deterministic Python checkers (no LLM judge); higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Claude Opus 4.6 | Thinking | 20 | #113 |
| 2 | Claude Sonnet 4.6 | Thinking | 14 | #173 |
| 3 | GPT-5.4 | High | 9 | #135 |
| 4 | Gemini 3.1 Pro (Preview) | 8 | #99 | |
| 5 | Kimi K2.6 | 7 | #192 | |
| 6 | Qwen 3.6 Plus | 5 | #238 | |
| 7 | Kimi K2.5 | 0 | #266 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=clawmark-task-success · How It Works · Data refreshed daily, snapshot 2026-11-21.