LoopsBench - Vendor Loops - Test Pass Rate (Single Segment): leaderboard
LoopsBench (arXiv 2608.00267v2, Table 2 within-family panel): 12 models on 112 dependency-structured long-horizon coding tasks, each run in its vendor's loop (GPT in Codex, Claude in Claude Code, Qwen in Qwen Code). Mean fraction of released tests passing without outer-loop continuation; percent, higher is better.
Metric: Released-test pass rate (%; without outer-loop continuation). Source: arxiv.org. Saturation forecast: Around July 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 41.18 | #51 |
| 2 | GPT-5.5 | 39.62 | #33 |
| 3 | Qwen 3.6 Plus | 37.97 | #124 |
| 4 | Claude Opus 4.5 | 36.97 | #87 |
| 5 | GPT-5.2 | 33.91 | #115 |
| 6 | Claude Sonnet 4.5 | 32.85 | #148 |
| 7 | Qwen 3.5 Plus | 31.75 | #133 |
| 8 | GPT-5 | 28.45 | #105 |
| 9 | Claude Sonnet 4 | 27.43 | #207 |
| 10 | GPT-4o | 21.85 | #357 |
| 11 | Qwen 3 Max | 17.45 | #213 |
| 12 | Qwen 2.5 72B | 10.61 | #512 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=loopsbench-vendor-loops-test-pass-rate-single-segment · How It Works · Data refreshed daily, snapshot 2026-11-05.