Grok 4 Tops with 94.15 Points: 2026-07-16 Smoke Quick Test Data Brief

The 2026-07-16 YZ Index Smoke Quick Test covered 9 models, with Grok 4 scoring 94.15 to take first place for the day. Smoke is a daily 10-question quick test, suitable for observing short-term signals, not equivalent to the Full weekly ranking conclusions.

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

Daily Ranking

Data Interpretation

In today's YZ Index Smoke Quick Test, Grok 4 topped the main ranking with 94.15, supported by a score structure of 100 for code execution and 87 for material constraints. DeepSeek V4 Pro followed closely with a main score of 93.84, also achieving a perfect 100 in code execution and 86.3 in material constraints, showing that both models achieved full marks in code execution while maintaining high scores in material constraints. Gemini 3.1 Pro scored 80.4 on the main ranking, with 75 for code execution and 87 for material constraints. Claude Opus 4.7 and Qwen3 Max both scored 80.09 on the main ranking, with 75 for code execution and 86.3 for material constraints. These three models exhibit a relatively balanced scoring pattern between code execution and material constraints.

Compared to the previous run with the same criteria, Gemini 2.5 Pro's main score dropped 41.5 points, code execution dropped 66.7 points, and material constraints dropped 10.7 points. Claude Sonnet 4.6's main score dropped 25.6 points, and code execution dropped 50 points. Claude Opus 4.7's main score dropped 19.9 points, code execution dropped 25 points, and material constraints dropped 13.7 points. Doubao Pro's main score dropped 15 points, code execution dropped 16.7 points, and material constraints dropped 13 points. Grok 4's main score rose 10.5 points, code execution rose 29.7 points, and material constraints dropped 13 points.

Claude Opus 4.7, Doubao Pro, Claude Sonnet 4.6, and Gemini 2.5 Pro experienced sharp declines in their main scores, possibly due to question sampling fluctuations or temporary model state changes, requiring subsequent runs for verification. GPT-5.5 and GPT-o3 had incomplete data due to API failures/timeouts and did not participate in this round's ranking.

Key Changes

• Gemini 2.5 Pro: Main score -41.5, Code execution -66.7, Material constraints -10.7
• Claude Sonnet 4.6: Main score -25.6, Code execution -50
• Claude Opus 4.7: Main score -19.9, Code execution -25, Material constraints -13.7
• Doubao Pro: Main score -15, Code execution -16.7, Material constraints -13
• Grok 4: Main score +10.5, Code execution +29.7, Material constraints -13

Signals to Watch

• Claude Opus 4.7: Main score sharp drop of -19.9 points
• Doubao Pro: Main score sharp drop of -15 points
• Claude Sonnet 4.6: Main score sharp drop of -25.6 points
• Gemini 2.5 Pro: Main score sharp drop of -41.5 points
• GPT-5.5: Incomplete data (missing execution, grounding, judgment, integrity, communication dimensions; API failure/timeout). Auto-retry triggered; not ranked in this round.
• GPT-o3: Incomplete data (missing execution, grounding, judgment, integrity, communication dimensions; API failure/timeout). Auto-retry triggered; not ranked in this round.

When reading such Smoke briefs, the focus should be on two questions: first, whether a model has repeatedly exposed the same weakness over multiple days; second, whether the integrity rating has changed from pass to warn or fail. Large daily fluctuations in execution or constraint scores may come from question sampling or could be early signals of actual degradation, requiring subsequent runs for verification.

Data source: YZ Index | Run #234 | View raw data