StructureClaw (Automatic Workflow): leaderboard
StructureClaw executable agent benchmark (arXiv 2607.14896); the automatic-workflow track reports the success rate of building a working structured-extraction agent without human guidance.
Metric: Success Rate (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.2 | 97.3 | #73 |
| 2 | GPT-5.5 | 91.3 | #31 |
| 3 | Qwen 3.7 Max | 89.3 | #75 |
| 4 | Gemini 3.5 Flash | 88.7 | #60 |
| 5 | Claude Opus 4.8 | 86.7 | #37 |
| 6 | DeepSeek V4 Flash | 83.3 | #137 |
| 7 | DeepSeek V4 Pro | 81.3 | #102 |
| 8 | Kimi K2.6 | 78 | #104 |
| 9 | MiniMax-M3 | 50 | #136 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=structureclaw-automatic-workflow · How It Works · Data refreshed daily, snapshot 2026-10-20.