LoopsBench - Claude Code Loop - Resolve Rate (Single Segment): 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. Share of tasks where every released test, regression obligations included, passes before any outer-loop continuation; percent, higher is better.
Metric: Resolve rate (%; without outer-loop continuation). Source: arxiv.org. Saturation forecast: Around September 2028. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 16.96 | #51 |
| 2 | GPT-5.5 | 13.39 | #33 |
| 3 | GLM-5.1 | 13.39 | #109 |
| 4 | DeepSeek V4 Pro | 11.61 | #112 |
| 5 | Gemini 3.1 Pro (Preview) | 8.93 | #64 |
| 6 | Qwen 3.6 Plus | 6.25 | #124 |
| 7 | Kimi K2.6 | 3.57 | #113 |
| 8 | Grok 4.1 Fast (Reasoning) | 2.68 | #227 (Grok 4.1 Fast) |
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-resolve-rate-single-segment · How It Works · Data refreshed daily, snapshot 2026-11-05.