GPT-o3 Tops with 80.61: 2026-07-18 Smoke Quick Test Data Brief

On 2026-07-18, the YZ Index Smoke quick test covered 11 models, with GPT-o3 ranking first at 80.61 points. Smoke is a daily 10-question quick test suitable for observing short-term signals and does not equate to the Full weekly ranking conclusions.

This Smoke evaluation only covers two main leaderboard dimensions: code execution and material constraints, with the leaderboard formula being 0.55 × Code Execution + 0.45 × Material Constraints. Due to the small daily sample size, single-day scores are more suitable as monitoring signals rather than drawing long-term conclusions about model capabilities.

Daily Ranking

Data Interpretation

In today's YZ Index Smoke quick test, the top models exhibited different pairings of code execution and material constraint scores. GPT-o3 led with a leaderboard score of 80.61, with code execution at 70.3 and material constraints at 93.2, showing a higher material constraint score. Claude Opus 4.7 had a leaderboard score of 79.1, with code execution at 75 and material constraints at 84.1, with relatively prominent code execution. Claude Sonnet 4.6 and GPT-5.5 had leaderboard scores of 74.6 and 74.1 respectively, both with code execution at 75, and material constraints between 74.1 and 73. Qwen3 Max had a leaderboard score of 70.15, with code execution at 64.7 and material constraints at 76.8, where material constraints provided some support. Overall, models with different strengths between code execution and material constraints occupied the top positions in the leaderboard.

Some models showed significant score changes compared to the previous comparable run. GLM-4.6 dropped 43.8 points in leaderboard, 91.7 points in code execution, and rose 14.8 points in material constraints. Gemini 3.1 Pro dropped 41.4 points in leaderboard and 76.6 points in code execution. Doubao Pro dropped 35.3 points in leaderboard and 61.7 points in code execution. Grok 4 dropped 25.1 points in leaderboard, 61.1 points in code execution, rose 19 points in material constraints, and its integrity changed from pass to warn. DeepSeek V4 Pro dropped 20.8 points in leaderboard, 26.3 points in code execution, and 14 points in material constraints. These changes may be due to question sampling fluctuations or may reflect real performance differences in the single-day test, requiring subsequent runs for verification.

The Smoke quick test is a small-sample single-day signal. The above observations are based solely on today's data and do not draw conclusions about long-term model performance. The specific score combinations of code execution and material constraints determine the leaderboard positions; abnormal changes require more comparable tests for verification.

Key Changes

• GLM-4.6: Leaderboard -43.8, Code Execution -91.7, Material Constraints +14.8
• Gemini 3.1 Pro: Leaderboard -41.4, Code Execution -76.6
• Doubao Pro: Leaderboard -35.3, Code Execution -61.7
• Grok 4: Leaderboard -25.1, Code Execution -61.1, Material Constraints +19, Integrity pass→warn
• DeepSeek V4 Pro: Leaderboard -20.8, Code Execution -26.3, Material Constraints -14

Signals to Watch

• No actionable abnormal signals were retained this time.

When reading such Smoke briefings, the focus should be on two questions: first, whether a model consistently exhibits the same weakness over multiple days; second, whether the integrity rating changes from pass to warn or fail. Large daily swings in execution or constraint scores may stem from question sampling or be early signals of genuine degradation, requiring subsequent runs for verification.

Data Source: YZ Index | Run #236 | View Raw Data