The 2026-11-18 YZ Index Smoke quick test covered 14 models, with GPT-6 Astra ranking first for the day at 91.14 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals, and is not equivalent to a Full weekly leaderboard conclusion.
This Smoke evaluation covered only two main leaderboard dimensions: code execution and material constraints. The main leaderboard formula is 0.55 × Code Execution + 0.45 × Material Constraints. Because the daily sample size is small, single-day scores are better used as monitoring signals rather than long-term conclusions about model capabilities.
Daily Ranking
| Rank | Model | Main Leaderboard | Code Execution | Material Constraints | Integrity |
|---|---|---|---|---|---|
| #1 | GPT-6 Astra | 91.14 | 100 | 80.3 | pass |
| #2 | GPT-6.1 Sol | 90.15 | 100 | 78.1 | pass |
| #3 | GPT-6 Luna | 83.91 | 75 | 94.8 | pass |
| #4 | Grok 4 | 83.9 | 79.4 | 89.4 | pass |
| #5 | Gemini 2.5 Pro | 82.52 | 75 | 91.7 | pass |
| #6 | DeepSeek V4 Pro | 81.48 | 75 | 89.4 | pass |
| #7 | GPT-5.5 | 78.74 | 75 | 83.3 | pass |
| #8 | GPT-6 Sol | 76.4 | 75 | 78.1 | pass |
| #9 | Claude Opus 4.7 | 75.99 | 70 | 83.3 | pass |
| #10 | GPT-o3 | 72.5 | 50 | 100 | pass |
| #11 | Doubao Pro | 71.81 | 75 | 67.9 | pass |
| #12 | GLM-4.6 | 71.51 | 50 | 97.8 | pass |
| #13 | Claude Sonnet 4.6 | 64.34 | 55.2 | 75.5 | pass |
| #14 | Gemini 3.1 Pro | 58.06 | 50 | 67.9 | pass |
Key Changes
- DeepSeek V4 Pro: Main leaderboard rose 42.2 points, Code Execution +50 points, Material Constraints +32.7 points
- Gemini 2.5 Pro: Main leaderboard rose 40.2 points, Code Execution +46.4 points, Material Constraints +32.6 points
- Grok 4: Main leaderboard rose 34.8 points, Code Execution +28.6 points, Material Constraints +42.4 points
- GLM-4.6: Main leaderboard rose 28.2 points, Code Execution +25 points, Material Constraints +32 points
- GPT-6 Astra: Main leaderboard rose 25.5 points, Code Execution +25.2 points, Material Constraints +25.9 points
Signals to Watch
- GPT-6 Luna: Code Execution plunged -24.1 points
- GPT-5.5: Code Execution plunged -24.8 points
- GPT-6 Sol: Code Execution plunged -24.1 points
- Claude Opus 4.7: Code Execution plunged -22.7 points
- GPT-o3: Code Execution plunged -16.3 points
- Claude Sonnet 4.6: Code Execution plunged -28.6 points
- Qwen3 Max: Data incomplete (multiple evaluation dimensions missing due to API failure/timeout); automatically queued for rerun and not included in this ranking
When reading this type of Smoke brief, focus on two questions: first, whether a model has exposed the same type of weakness for multiple consecutive days; second, whether its integrity rating has moved from pass to warn or fail. Large single-day changes in execution or constraint scores may come from question sampling or may be an early signal of real degradation, and require follow-up runs to review.
Data source: YZ Index | Run #432 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接