On 2026-10-13, the YZ Index Smoke quick test covered 13 models, with Gemini 3.1 Pro ranking first for the day at 82.39 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals; it is not equivalent to the conclusions of the Full weekly leaderboard.
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. Due to the small daily sample size, 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 | Gemini 3.1 Pro | 82.39 | 91.7 | 71 | pass |
| #2 | GPT-o3 | 70.91 | 75 | 65.9 | pass |
| #3 | Grok 4 | 70.91 | 75 | 65.9 | warn |
| #4 | Doubao Pro | 69.11 | 75 | 61.9 | pass |
| #5 | DeepSeek V4 Pro | 64.27 | 50 | 81.7 | pass |
| #6 | GPT-6 Luna | 62.85 | 75 | 48 | pass |
| #7 | GPT-6.1 Sol | 61 | 58.3 | 64.3 | pass |
| #8 | Claude Sonnet 4.6 | 60.17 | 50 | 72.6 | pass |
| #9 | GPT-5.5 | 59.92 | 58.3 | 61.9 | pass |
| #10 | GPT-6 Astra | 59.92 | 58.3 | 61.9 | pass |
| #11 | GPT-6 Sol | 59.84 | 75 | 41.3 | pass |
| #12 | Claude Opus 4.7 | 59.45 | 50 | 71 | pass |
| #13 | Gemini 2.5 Pro | 57.16 | 50 | 65.9 | pass |
Data Interpretation
In today's YZ Index Smoke quick test, Gemini 3.1 Pro ranked first with a main leaderboard score of 82.39; its 91.7 in code execution and 71 in material constraints form a clearly strong combination, and under the 0.55 weight, its code execution contribution stands out. GPT-o3 and Grok 4 tied at 70.91 on the main leaderboard, both showing a structure of 75 in code execution and 65.9 in material constraints; leading models overall rely on score momentum from the code execution side. DeepSeek V4 Pro is characterized by 81.7 in material constraints, and its main leaderboard score of 64.27 reflects a relative advantage on the material constraints side.
GPT-6 Astra fell 25.1 points on the main leaderboard, 38.7 in code execution, and 8.4 in material constraints; GPT-6.1 Sol fell 24 points on the main leaderboard, 38.7 in code execution, and 6 in material constraints; GPT-6 Sol fell 23 points on the main leaderboard, 19.8 in code execution, and 26.8 in material constraints; GPT-6 Luna fell 15.7 points on the main leaderboard, 22 in code execution, and 8 in material constraints; and Doubao Pro fell 15.6 points on the main leaderboard, 16.5 in code execution, and 14.4 in material constraints. All are significant single-day movements. These changes may stem from question sampling fluctuations or may reflect real degradation, requiring subsequent same-methodology run verification.
The Smoke quick test is a small-sample single-day signal. Current data only shows different combinations between code execution and material constraints among leading models, with no other anomalies observed. Interpretation remains restrained to avoid over-inference.
Main Changes
- GPT-6 Astra: main leaderboard down 25.1 points, code execution -38.7, material constraints -8.4
- GPT-6.1 Sol: main leaderboard down 24 points, code execution -38.7, material constraints -6
- GPT-6 Sol: main leaderboard down 23 points, code execution -19.8, material constraints -26.8
- GPT-6 Luna: main leaderboard down 15.7 points, code execution -22, material constraints -8
- Doubao Pro: main leaderboard down 15.6 points, code execution -16.5, material constraints -14.4
Signals to Watch
- No publishable anomalous signals were retained this time.
When reading this type of Smoke brief, focus on two questions: first, whether a model has repeatedly exposed the same type of weakness over 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 early signals of real degradation, requiring subsequent run verification.
Data source: YZ Index | Run #375 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接