On 2026-10-14, the YZ Index Smoke quick test covered 13 models, with Claude Opus 4.7, DeepSeek V4 Pro, and GPT-o3 tied for first place that day at 92.39. Smoke is a daily 10-question quick test, suitable for observing short-term signals; it is not equivalent to a Full weekly leaderboard conclusion.
This Smoke evaluation covers 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, a single-day score is better used as a monitoring signal rather than a long-term verdict on model capability.
Daily Ranking
| Rank | Model | Main Leaderboard | Code Execution | Material Constraints | Integrity |
|---|---|---|---|---|---|
| #1 | Claude Opus 4.7 | 92.39 | 94.5 | 89.8 | pass |
| #2 | DeepSeek V4 Pro | 92.39 | 94.5 | 89.8 | pass |
| #3 | GPT-o3 | 92.39 | 94.5 | 89.8 | pass |
| #4 | GPT-5.5 | 88.6 | 89.5 | 87.5 | pass |
| #5 | Doubao Pro | 88.54 | 85.3 | 92.5 | pass |
| #6 | GPT-6.1 Sol | 88.29 | 89.5 | 86.8 | pass |
| #7 | GPT-6 Astra | 87.25 | 89.5 | 84.5 | pass |
| #8 | GPT-6 Sol | 86.81 | 88.7 | 84.5 | pass |
| #9 | Grok 4 | 83.2 | 77.8 | 89.8 | warn |
| #10 | Claude Sonnet 4.6 | 78.64 | 69.5 | 89.8 | pass |
| #11 | Gemini 2.5 Pro | 76.33 | 65.3 | 89.8 | pass |
| #12 | Gemini 3.1 Pro | 75.89 | 64.5 | 89.8 | pass |
| #13 | GPT-6 Luna | 74.54 | 64.5 | 86.8 | pass |
Data Interpretation
Today's YZ Index Smoke quick test shows that Claude Opus 4.7, DeepSeek V4 Pro, and GPT-o3 are tied on the main leaderboard at 92.39, each with Code Execution at 94.5 and Material Constraints at 89.8; the three display a balanced combination of strengths across Code Execution and Material Constraints. GPT-5.5 has a main leaderboard score of 88.6, Code Execution of 89.5, and Material Constraints of 87.5; Doubao Pro has a main leaderboard score of 88.54, Code Execution of 85.3, and Material Constraints of 92.5, with its Material Constraints relatively prominent. GPT-6.1 Sol has a main leaderboard score of 88.29, Code Execution of 89.5, and Material Constraints of 86.8; such models maintain a certain balance between the two dimensions.
Compared with the previous same-methodology run, Claude Opus 4.7 rose by +32.9 points on the main leaderboard, +44.5 in Code Execution, and +18.8 in Material Constraints; DeepSeek V4 Pro rose by +28.1 on the main leaderboard, +44.5 in Code Execution, and +8.1 in Material Constraints; GPT-5.5 rose by +28.7 on the main leaderboard, +31.2 in Code Execution, and +25.6 in Material Constraints. These notable movers mainly reflect large gains in the Code Execution dimension. GPT-6.1 Sol rose by +27.3 on the main leaderboard, +31.2 in Code Execution, and +22.5 in Material Constraints, while GPT-6 Astra rose by +27.3 on the main leaderboard, +31.2 in Code Execution, and +22.6 in Material Constraints, indicating synchronous changes in Code Execution and Material Constraints within the score structure.
Gemini 3.1 Pro's Code Execution dropped sharply by -27.2 points; this anomalous signal may stem from question sampling fluctuation or genuine regression and requires confirmation in a follow-up run. GLM-4.6 and Qwen3 Max did not participate in the ranking due to incomplete data. As Smoke is a small-sample single-day signal, related interpretations should remain measured.
Key Changes
- Claude Opus 4.7: Main leaderboard up 32.9 points, Code Execution +44.5, Material Constraints +18.8
- GPT-5.5: Main leaderboard up 28.7 points, Code Execution +31.2, Material Constraints +25.6
- DeepSeek V4 Pro: Main leaderboard up 28.1 points, Code Execution +44.5, Material Constraints +8.1
- GPT-6.1 Sol: Main leaderboard up 27.3 points, Code Execution +31.2, Material Constraints +22.5
- GPT-6 Astra: Main leaderboard up 27.3 points, Code Execution +31.2, Material Constraints +22.6
Signals to Watch
- Gemini 3.1 Pro: Code Execution plunges -27.2 points
- GLM-4.6: Incomplete data (some evaluation dimensions missing due to API failure/timeout); an automatic rerun has been initiated; not included in this period's ranking
- Qwen3 Max: Incomplete data (multiple evaluation dimensions missing due to API failure/timeout); an automatic rerun has been initiated; not included in this period's ranking
When reading this kind of Smoke briefing, the focus should be 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 an early signal of genuine regression, requiring review in a follow-up run.
Data source: YZ Index | Run #378 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接