LoopArena - Task Slice: leaderboard
Part of LoopArena (Alibaba DreamX, 2026): the evaluated model controls a fixed Qwen3.7-Plus coding agent through 27 condensed slices of long coding tasks, deciding its next instruction or when to stop. Strict success rate in percent over three runs per slice; higher is better.
Metric: Strict success rate (%; 27 task slices paired with the full tasks (Type II); the evaluated model is the Controller: after each round it reads a Reporter's evidence packet and issues the next instruction to, or stops, a fixed Qwen3.7-Plus coding Worker, which alone has coding tools; a run succeeds only when it passes the task evaluator (every Core check for SCBench, reward 1 from the official Harbor evaluator for BeyondSWE) and follows the control protocol; three runs per task; provider-default reasoning). Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 51.85 | #49 |
| 2 | Claude Opus 4.8 | 48.15 | #56 |
| 3 | Qwen 3.7 Plus | 48.15 | #207 |
| 4 | DeepSeek V4 Flash (0731) | 45.68 | #139 |
| 5 | GLM-5.2 | 37.04 | #138 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=looparena-task-slice · How It Works · Data refreshed daily, snapshot 2026-11-13.