Cloud-OpsBench (OnlineBoutique) - Component Accuracy: leaderboard
Metric: Component accuracy (%): share of episodes whose predicted faulty component matches the label, on Cloud-OpsBench's OnlineBoutique fault cases (part of 754 runtime-verified cases over 57 fault types); the model drives the authors' ReAct agent (Pydantic-validated tools, temperature 0, at most 20 steps, extended thinking disabled) over snapshot-replayed Kubernetes diagnostic tools, from a user query describing the observed failure; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Flash (Non-reasoning) | 88 | #134 (DeepSeek V4 Flash) |
| 2 | GPT-5 | 86 | #91 |
| 3 | Qwen 3.5 27B (Non-reasoning) | 85 | #183 (Qwen 3.5 27B) |
| 4 | Gemini 2.5 Flash (Non-reasoning) | 82 | #237 (Gemini 2.5 Flash) |
| 5 | Gemini 2.5 Pro | 81 | #145 |
| 6 | Qwen 3.5 Plus (Non-reasoning) | 80 | #123 (Qwen 3.5 Plus) |
| 7 | Claude Sonnet 4 | 71 | #194 |
| 8 | Qwen 3 14B (Non-reasoning) | 61 | #524 (Qwen 3 14B) |
| 9 | Qwen 3 235B A22B (Non-reasoning) | 61 | #304 (Qwen 3 235B A22B) |
| 10 | Qwen 3 8B (Non-reasoning) | 35 | #667 (Qwen 3 8B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=cloud-opsbench-onlineboutique-component-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.