The 2026-11-07 YZ Index Smoke quick test covered 14 models, with GPT-6 Astra ranking first for the day at 93.05 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 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, single-day scores are better used as monitoring signals rather than long-term conclusions about model capability.
Daily Ranking
| Rank | Model | Main Leaderboard | Code Execution | Material Constraints | Integrity |
|---|---|---|---|---|---|
| #1 | GPT-6 Astra | 93.05 | 99.8 | 84.8 | pass |
| #2 | GPT-6 Luna | 86.98 | 99.8 | 71.3 | pass |
| #3 | GPT-6 Sol | 86.17 | 99.8 | 69.5 | pass |
| #4 | GPT-6.1 Sol | 86.17 | 99.8 | 69.5 | pass |
| #5 | GPT-o3 | 82.92 | 91.6 | 72.3 | pass |
| #6 | GPT-5.5 | 82.48 | 99.8 | 61.3 | pass |
| #7 | Doubao Pro | 79.57 | 96.8 | 58.5 | pass |
| #8 | Gemini 3.1 Pro | 79.28 | 75 | 84.5 | pass |
| #9 | Claude Opus 4.7 | 78.91 | 70 | 89.8 | pass |
| #10 | Claude Sonnet 4.6 | 75.8 | 89.3 | 59.3 | pass |
| #11 | Grok 4 | 72.8 | 71.4 | 74.5 | pass |
| #12 | Gemini 2.5 Pro | 64.31 | 50 | 81.8 | pass |
| #13 | GLM-4.6 | 64.31 | 50 | 81.8 | pass |
| #14 | DeepSeek V4 Pro | 63.28 | 50 | 79.5 | pass |
Data Interpretation
In today's Smoke quick test, GPT-6 Astra led the main leaderboard at 93.05, with a strong combination of 99.8 in Code Execution and 84.8 in Material Constraints creating a clear advantage. GPT-6 Luna and GPT-6 Sol both scored 99.8 in Code Execution, with Material Constraints at 71.3 and 69.5 and main leaderboard scores of 86.98 and 86.17, respectively. Claude Opus 4.7 had a relatively high Material Constraints score of 89.8, but its Code Execution was 70 and its main leaderboard score was 78.91, showing a structure with strong Material Constraints and relatively weak Code Execution. Gemini 3.1 Pro had Code Execution of 75 and Material Constraints of 84.5, with a main leaderboard score of 79.28, also reflecting how Material Constraints supported its ranking.
GPT-6 Luna's main leaderboard score rose 22.9 points from the previous run, with Code Execution up 49.8 points and Material Constraints down 10.1 points; GPT-6 Sol's main leaderboard score rose 18.1 points, with Code Execution up 30.3 points; GLM-4.6's main leaderboard score rose 15.7 points, with Code Execution up 30.5 points; GPT-o3's main leaderboard score rose 15.1 points, with Code Execution up 22.1 points and Material Constraints up 6.6 points; GPT-6 Astra's main leaderboard score rose 13.8 points, with Material Constraints up 24.1 points. These shifts may stem from single-day question sampling variance and require follow-up runs to verify whether they reflect real changes.
In terms of anomalous signals, GPT-5.5's Material Constraints fell 18 points, Doubao Pro's main leaderboard score fell 8 points, Gemini 2.5 Pro's main leaderboard score fell 9.5 points, and DeepSeek V4 Pro's main leaderboard score fell 10.2 points. Qwen3 Max was not ranked due to incomplete data. The Smoke test is a small-sample, single-day signal; the above fluctuations require multiple runs to verify stability and should not be used for long-term judgments.
Key Changes
- GPT-6 Luna: main leaderboard score up 22.9 points, Code Execution +49.8 points, Material Constraints -10.1 points
- GPT-6 Sol: main leaderboard score up 18.1 points, Code Execution +30.3 points
- GLM-4.6: main leaderboard score up 15.7 points, Code Execution +30.5 points
- GPT-o3: main leaderboard score up 15.1 points, Code Execution +22.1 points, Material Constraints +6.6 points, Integrity warn→pass
- GPT-6 Astra: main leaderboard score up 13.8 points, Code Execution +5.3 points, Material Constraints +24.1 points
Signals to Watch
- GPT-5.5: Material Constraints plunged -18 points
- Doubao Pro: main leaderboard score plunged -8 points
- Gemini 2.5 Pro: main leaderboard score plunged -9.5 points
- DeepSeek V4 Pro: main leaderboard score plunged -10.2 points
- Qwen3 Max: incomplete data (missing execution, judgment, integrity, and communication dimensions; API failure/timeout); it has entered automatic rerun and is not ranked in this issue
When reading this kind of Smoke brief, focus on two questions: first, whether a model has exposed the same type of weakness for several 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, and require follow-up runs for review.
Data source: YZ Index | Run #415 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接