On 2026-10-25, the YZ Index Smoke quick test covered 14 models, with Grok 4 ranking first for the day at 96.65 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals and not equivalent to Full weekly ranking conclusions.
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 | Grok 4 | 96.65 | 93.9 | 100 | pass |
| #2 | GPT-5.5 | 95.19 | 100 | 89.3 | pass |
| #3 | GPT-6 Luna | 94.09 | 98 | 89.3 | pass |
| #4 | Doubao Pro | 93.48 | 96.9 | 89.3 | pass |
| #5 | GPT-6 Sol | 88.68 | 92.5 | 84 | pass |
| #6 | GPT-6 Astra | 85.44 | 94.8 | 74 | pass |
| #7 | Claude Opus 4.7 | 81.44 | 75 | 89.3 | pass |
| #8 | Gemini 3.1 Pro | 81.44 | 75 | 89.3 | pass |
| #9 | GPT-6.1 Sol | 79.82 | 94.8 | 61.5 | pass |
| #10 | GPT-o3 | 76.8 | 70.5 | 84.5 | pass |
| #11 | Claude Sonnet 4.6 | 72.5 | 50 | 100 | pass |
| #12 | Gemini 2.5 Pro | 71.29 | 47.8 | 100 | pass |
| #13 | DeepSeek V4 Pro | 67.69 | 50 | 89.3 | pass |
| #14 | GLM-4.6 | 31.44 | 25 | 39.3 | pass |
Data Interpretation
In today's YZ Index Smoke quick test, the top models showed clear differentiation in the pairing of code execution and material constraints. Grok 4 ranked first with a main leaderboard score of 96.65, with code execution 93.9 and material constraints 100; GPT-5.5 had a main leaderboard score of 95.19, with code execution 100 and material constraints 89.3; GPT-6 Luna had a main leaderboard score of 94.09, with code execution 98 and material constraints 89.3. All three passed the integrity check, showing that the combination of high code execution and high material constraints formed a leading advantage in the day's small sample.
Some models showed strength in a single dimension but were limited overall. Claude Sonnet 4.6 had a main leaderboard score of 72.5, with code execution 50 and material constraints 100; Gemini 2.5 Pro had a main leaderboard score of 71.29, with code execution 47.8 and material constraints 100; GLM-4.6 had a main leaderboard score of 31.44, with code execution 25 and material constraints 39.3. Such pairings scored relatively low under the main leaderboard formula, reflecting a structural feature in the day's test of high material constraints but low code execution.
Qwen3 Max had incomplete data due to an API failure, with some evaluation dimensions unavailable, and did not participate in this period's ranking. This signal may stem from sampling variation in the questions or a temporary interface issue and needs follow-up review to confirm whether it reflects real degradation. The Smoke quick test is a single-day, small-sample signal; the observations above are for reference only for that day.
Key Changes
- No historical test under the same methodology is available for comparison, so only the day's data is shown.
Signals to Watch
- Qwen3 Max: Incomplete data (some evaluation dimensions unavailable due to API failure/timeout), automatically scheduled for a follow-up test, and not included in this period's ranking
When reading this kind of Smoke brief, focus on two questions: first, whether a model has repeatedly shown the same type of weakness over multiple days; second, whether its integrity rating moves 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 real degradation, requiring follow-up review.
Data source: YZ Index | Run #394 | View raw data
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接