On 2026-10-29, the YZ Index Smoke quick test covered 14 models. Claude Sonnet 4.6, GPT-6 Astra, GPT-6 Sol, and GPT-6.1 Sol tied for first place of the day with 87.36 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals; it is not equivalent to conclusions from the Full weekly ranking.
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 capability.
Daily Ranking
| Rank | Model | Overall | Code Execution | Material Constraints | Integrity |
|---|---|---|---|---|---|
| #1 | Claude Sonnet 4.6 | 87.36 | 100 | 71.9 | pass |
| #2 | GPT-6 Astra | 87.36 | 100 | 71.9 | pass |
| #3 | GPT-6 Sol | 87.36 | 100 | 71.9 | pass |
| #4 | GPT-6.1 Sol | 87.36 | 100 | 71.9 | pass |
| #5 | Doubao Pro | 85.11 | 100 | 66.9 | pass |
| #6 | GPT-o3 | 78.42 | 75 | 82.6 | pass |
| #7 | Gemini 3.1 Pro | 77.46 | 82 | 71.9 | pass |
| #8 | Claude Opus 4.7 | 73.61 | 75 | 71.9 | pass |
| #9 | DeepSeek V4 Pro | 73.61 | 75 | 71.9 | pass |
| #10 | Gemini 2.5 Pro | 71.3 | 70.8 | 71.9 | pass |
| #11 | Grok 4 | 71.19 | 82.3 | 57.6 | pass |
| #12 | GPT-5.5 | 70.59 | 75 | 65.2 | pass |
| #13 | GPT-6 Luna | 67.17 | 75 | 57.6 | pass |
| #14 | GLM-4.6 | 59.86 | 50 | 71.9 | pass |
Data Interpretation
In today's YZ Index Smoke quick test, the top models showed clear patterns in the combination of Code Execution and Material Constraints. Claude Sonnet 4.6, GPT-6 Astra, GPT-6 Sol, and GPT-6.1 Sol all scored 87.36 on the main leaderboard, with Code Execution at 100 and Material Constraints at 71.9; this structure of high Code Execution paired with moderate Material Constraints placed them at the front. Doubao Pro scored 85.11 on the main leaderboard, with Code Execution at 100 and Material Constraints at 66.9, also relying on its Code Execution advantage. By contrast, GPT-o3 scored 78.42 on the main leaderboard, with Code Execution at 75 and Material Constraints at 82.6, showing a stronger Material Constraints combination; Gemini 3.1 Pro scored 77.46 on the main leaderboard, with Code Execution at 82 and Material Constraints at 71.9, relatively balanced across the two dimensions.
Compared with the previous run under the same methodology, some models showed significant changes. Gemini 3.1 Pro rose 31.8 points on the main leaderboard and 57 points in Code Execution; DeepSeek V4 Pro rose 27.9 points on the main leaderboard and 50 points in Code Execution; Claude Sonnet 4.6 rose 20.7 points on the main leaderboard, 25.2 points in Code Execution, and 15.2 points in Material Constraints; Gemini 2.5 Pro rose 14.9 points on the main leaderboard, 20.8 points in Code Execution, and 7.6 points in Material Constraints. These changes were concentrated in the Code Execution dimension, reflecting the impact of that dimension's single-day fluctuation on the main leaderboard.
In terms of anomalous signals, Grok 4's Material Constraints plunged by 17.4 points; GPT-6 Luna's main leaderboard score plunged by 19.8 points, with Code Execution down 25 points and Material Constraints down 13.4 points. Qwen3 Max did not participate in the ranking because of incomplete data. The above changes may stem from question sampling fluctuation or may reflect real degradation, and require follow-up runs for confirmation. The Smoke quick test is a small-sample, single-day signal; the above observations are for reference only.
Key Changes
- Gemini 3.1 Pro: main leaderboard up 31.8 points, Code Execution +57 points
- DeepSeek V4 Pro: main leaderboard up 27.9 points, Code Execution +50 points
- Claude Sonnet 4.6: main leaderboard up 20.7 points, Code Execution +25.2 points, Material Constraints +15.2 points
- GPT-6 Luna: main leaderboard down 19.8 points, Code Execution -25 points, Material Constraints -13.4 points
- Gemini 2.5 Pro: main leaderboard up 14.9 points, Code Execution +20.8 points, Material Constraints +7.6 points
Signals to Watch
- Grok 4: Material Constraints plunged by 17.4 points
- GPT-6 Luna: main leaderboard plunged by 19.8 points
- Qwen3 Max: data incomplete due to API failure/timeout, has entered automatic rerun, and is not included in this issue's ranking
When reading this kind of Smoke brief, the focus should be on two questions: first, whether a model has repeatedly exposed the same type of weakness over multiple days; second, whether the Integrity rating has moved from pass to warn or fail. Large single-day changes in Code Execution or Material Constraints scores may come from question sampling or may be early signals of real degradation, and require follow-up runs for confirmation.
Data source: YZ Index | Run #401 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接