SAKE - Zero-Shot - Overall: leaderboard
Evaluates software architectural knowledge through multiple-choice questions on the paper's 2,114 scored test items, with temperature zero and short answer limits. The paper reports model display labels without exact endpoint snapshots; those labels are retained. This series reports SAKE - Zero-Shot - Overall using Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 93.95 | #83 |
| 2 | Claude Sonnet 4.6 | 93.71 | #94 |
| 3 | Gemini 3.1 Flash Lite | 93.66 | #200 |
| 4 | DeepSeek V3.2 | 93.61 | #210 |
| 5 | Claude Opus 4.6 | 93.57 | #67 |
| 6 | Qwen 3 235B A22B | 93.38 | #323 |
| 7 | Grok 4.20 | 93.28 | #179 |
| 8 | Mistral Small 4 | 92.48 | #522 |
| 9 | Claude Haiku 4.5 | 92.15 | #287 |
| 10 | Grok 4.1 Fast | 92.05 | #227 |
| 11 | GPT-5.4 Nano | 90.96 | #359 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sake-zero-shot-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.