2026-07-14 YZ Index Smoke Quick Test covered 11 models, with DeepSeek V4 Pro ranking first on the day with 91.46 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals, not equivalent to Full weekly rankings.
This Smoke evaluation only covers two main dimensions: Code Execution and Material Constraint. The main ranking formula is 0.55 × Code Execution + 0.45 × Material Constraint. Due to the small daily sample size, single-day scores are more suitable as monitoring signals rather than definitive long-term conclusions about model capabilities.
Daily Ranking
| Rank | Model | Main Score | Code Execution | Material Constraint | Integrity |
|---|---|---|---|---|---|
| #1 | DeepSeek V4 Pro | 91.46 | 98.7 | 82.6 | pass |
| #2 | GPT-5.5 | 91.32 | 100 | 80.7 | pass |
| #3 | Claude Opus 4.7 | 86.25 | 75 | 100 | pass |
| #4 | Gemini 2.5 Pro | 86.1 | 100 | 69.1 | pass |
| #5 | GPT-o3 | 86.1 | 100 | 69.1 | pass |
| #6 | Gemini 3.1 Pro | 80.99 | 75 | 88.3 | pass |
| #7 | Doubao Pro | 78.42 | 75 | 82.6 | warn |
| #8 | Grok 4 | 74.91 | 75 | 74.8 | warn |
| #9 | Qwen3 Max | 62.42 | 58.5 | 67.2 | pass |
| #10 | Claude Sonnet 4.6 | 61.1 | 75 | 44.1 | pass |
| #11 | GLM-4.6 | 45.63 | 62.5 | 25 | warn |
Data Interpretation
In today's YZ Index Smoke Quick Test, DeepSeek V4 Pro ranked first with a main score of 91.46, with a balanced combination of Code Execution (98.7) and Material Constraint (82.6); GPT-5.5 followed closely with a main score of 91.32, achieving a perfect Code Execution score of 100 but a lower Material Constraint of 80.7, highlighting its code execution advantage. Claude Opus 4.7, with a main score of 86.25, was driven by a perfect Material Constraint score of 100, while its Code Execution score of 75 was relatively low, reflecting a structure dominated by material constraint. Gemini 2.5 Pro and GPT-o3 tied at 86.1 on the main score, both showing a combination of Code Execution 100 and Material Constraint 69.1.
Several models experienced significant declines: Grok 4's main score dropped by 22.5 points, with Code Execution and Material Constraint falling by 22 and 23 points respectively; GLM-4.6's main score dropped by 21.2 points, with Code Execution down 19.8 points and Material Constraint down 22.8 points; Claude Sonnet 4.6 and Doubao Pro saw main score drops of 18.5 and 18.3 points respectively; Gemini 3.1 Pro's main score fell by 15.7 points. These changes may result from question sampling fluctuations or could indicate actual single-day performance degradation, requiring subsequent consistent run reviews for confirmation.
The Smoke Quick Test is a small-sample single-day signal; interpretation is based solely on today's data and should not be used for long-term inferences. Models with a integrity rating of warn include Doubao Pro, Grok 4, and GLM-4.6; all other models passed.
Key Changes
- Grok 4: Main score -22.5, Code Execution -22, Material Constraint -23, Integrity pass→warn
- GLM-4.6: Main score -21.2, Code Execution -19.8, Material Constraint -22.8
- Claude Sonnet 4.6: Main score -18.5, Code Execution -19, Material Constraint -17.9
- Doubao Pro: Main score -18.3, Code Execution -19, Material Constraint -17.4, Integrity pass→warn
- Gemini 3.1 Pro: Main score -15.7, Code Execution -19, Material Constraint -11.7
Signals to Watch
- No reportable anomaly signals were retained in this test.
When reading such Smoke briefs, the focus should be on two questions: first, whether a model has repeatedly exposed the same weakness over consecutive days; second, whether the integrity rating has shifted from pass to warn or fail. Significant daily changes in execution or constraint scores may stem from question sampling or could be early signals of actual degradation, requiring subsequent run reviews.
Data Source: YZ Index | Run #231 | View Raw Data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接