On 2026-07-11, the YZ Index Smoke quick test covered 11 models. Claude Opus 4.7, Claude Sonnet 4.6, and GPT-o3 tied for first place that day with 81.44 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals, and is not equivalent to the Full weekly ranking conclusions.
This Smoke evaluation only covered two main board dimensions: code execution and material constraints. The main board formula is 0.55 × code execution + 0.45 × material constraints. Due to the small daily sample size, daily scores are more suitable as monitoring signals rather than long-term conclusions about model capabilities.
Daily Rankings
| Rank | Model | Main Board | Code Execution | Material Constraints | Integrity |
|---|---|---|---|---|---|
| #1 | Claude Opus 4.7 | 81.44 | 75 | 89.3 | pass |
| #2 | Claude Sonnet 4.6 | 81.44 | 75 | 89.3 | pass |
| #3 | GPT-o3 | 81.44 | 75 | 89.3 | pass |
| #4 | Doubao Pro | 80.78 | 73.8 | 89.3 | warn |
| #5 | GPT-5.5 | 75 | 75 | 75 | pass |
| #6 | Grok 4 | 74.56 | 62.5 | 89.3 | pass |
| #7 | Qwen3 Max | 74.56 | 62.5 | 89.3 | pass |
| #8 | DeepSeek V4 Pro | 67.69 | 50 | 89.3 | pass |
| #9 | Gemini 3.1 Pro | 67.69 | 50 | 89.3 | pass |
| #10 | Gemini 2.5 Pro | 60.81 | 37.5 | 89.3 | pass |
| #11 | GLM-4.6 | 36.25 | 25 | 50 | warn |
Data Interpretation
The top three models on today's main board—Claude Opus 4.7, Claude Sonnet 4.6, and GPT-o3—all scored 81.44 points, with code execution at 75 and material constraints at 89.3, indicating a balanced combination of these two capabilities. Doubao Pro closely followed with 80.78 points, scoring 73.8 in code execution and 89.3 in material constraints—matching the top three in material constraints while trailing slightly in code execution. GLM-4.6, ranking lower, scored only 36.25 on the main board, with code execution at 25 and material constraints at 50, both in a low range.
Compared to the previous run under the same settings, GLM-4.6's main board score dropped by 22.9 points, code execution fell by 69.5 points, and material constraints rose by 34.1 points; Gemini 2.5 Pro's main board dropped by 19.7 points, code execution fell by 56.2 points, and material constraints rose by 25 points; Gemini 3.1 Pro's main board dropped by 16.9 points, code execution fell by 43.7 points, and material constraints rose by 15.9 points. Claude Sonnet 4.6's main board rose by 15.5 points, code execution increased by 5.5 points, and material constraints rose by 27.6 points. These changes may result from daily question sampling fluctuations, or may reflect actual model performance differences, requiring verification in subsequent runs. The Smoke test provides small-sample daily signals; the above observations are for reference on this day only.
Key Changes
- GLM-4.6: Main board -22.9, code execution -69.5, material constraints +34.1
- Gemini 2.5 Pro: Main board -19.7, code execution -56.2, material constraints +25
- Gemini 3.1 Pro: Main board -16.9, code execution -43.7, material constraints +15.9
- Claude Sonnet 4.6: Main board +15.5, code execution +5.5, material constraints +27.6
- DeepSeek V4 Pro: Main board -14.2, code execution -40.9, material constraints +18.5
Signals to Watch
- No publishable anomaly signals were retained in this run.
When reading such Smoke briefs, the focus should be on two questions: first, whether a particular model has shown the same weakness for multiple consecutive days; second, whether the integrity rating has dropped from pass to warn or fail. Large daily fluctuations in execution or constraint scores may stem from question sampling or could be early signals of real degradation, requiring verification in subsequent runs.
Data source: YZ Index | Run #225 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接