LoopsBench - Claude Code Loop - Dependency Depth: leaderboard
LoopsBench (arXiv 2608.00267v2, Table 2 model sweep): eight models from different vendors on 112 dependency-structured long-horizon coding tasks, all run in the Claude Code loop. Mean fraction of the task's prerequisite layers fully passing before the first residual handoff; 0-1, higher is better.
Metric: Dependency depth reached before the first residual handoff (0-1). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 0.61 | #51 |
| 2 | GPT-5.5 | 0.47 | #33 |
| 3 | DeepSeek V4 Pro | 0.46 | #112 |
| 4 | GLM-5.1 | 0.44 | #109 |
| 5 | Qwen 3.6 Plus | 0.34 | #124 |
| 6 | Gemini 3.1 Pro (Preview) | 0.31 | #64 |
| 7 | Grok 4.1 Fast (Reasoning) | 0.15 | #227 (Grok 4.1 Fast) |
| 8 | Kimi K2.6 | 0.14 | #113 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=loopsbench-claude-code-loop-dependency-depth · How It Works · Data refreshed daily, snapshot 2026-11-05.