Claude Opus 4.7 Leads with 83.48 Points: 2026-11-14 Smoke Quick Test Data Brief

On 2026-11-14, the YZ Index Smoke Quick Test covered 13 models, and Claude Opus 4.7 ranked first for the day with 83.48 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals and not equivalent to conclusions from the Full weekly leaderboard.

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

Data Interpretation

In terms of score structure, Claude Opus 4.7 ranks first with a Main Leaderboard score of 83.48; its Code Execution score of 94.5 and Material Constraints score of 70 form a relatively balanced combination, and its Material Constraints score is the highest among the top models. GPT-6 Luna, GPT-5.5, Doubao Pro, GPT-6.1 Sol, and Claude Sonnet 4.6 all show a high Code Execution score of 94.5, while their Material Constraints scores remain at 54.8, 54.3, 47.6, 44.8, and 41.4, respectively, showing a common pattern of strong Code Execution and relatively weak Material Constraints. Grok 4 and GPT-6 Sol have Code Execution scores of 88.8 and 91.3, Material Constraints scores of 44.8 and 34.8, and overall Main Leaderboard scores of 69 and 65.88.

Among anomalous signals, GPT-6 Astra's Main Leaderboard score dropped sharply by 27.3 points, accompanied by a 60.2-point drop in Material Constraints; GPT-6 Luna's Main Leaderboard score rose by 16.3 points, accompanied by a 35-point rise in Material Constraints; Doubao Pro's Main Leaderboard score rose by 15.8 points, accompanied by a 19.5-point rise in Code Execution and an 11.3-point rise in Material Constraints; GPT-o3's Main Leaderboard score fell by 14.2 points, accompanied by a 27.2-point drop in Material Constraints; and Claude Sonnet 4.6's Main Leaderboard score fell by 11.7 points, accompanied by a 5-point drop in Code Execution and a 19.9-point drop in Material Constraints. These large single-day fluctuations may stem from question sampling variance or may reflect real degradation, and follow-up runs are needed to confirm. Claude Opus 4.7's Main Leaderboard score plunging by 11.4 points, Grok 4's Material Constraints plunging by 16.5 points, Gemini 2.5 Pro's Material Constraints plunging by 16.5 points, and Gemini 3.1 Pro's Code Execution plunging by 30.5 points likewise highlight the limitations of small-sample signals.

GLM-4.6 and Qwen3 Max did not participate in the ranking due to incomplete data and have entered automatic rerun; the results of this Smoke Quick Test are for same-day reference only.

Key Changes

• GPT-6 Astra: Main Leaderboard down 27.3 points; Material Constraints -60.2 points
• GPT-6 Luna: Main Leaderboard up 16.3 points; Material Constraints +35 points
• Doubao Pro: Main Leaderboard up 15.8 points; Code Execution +19.5 points; Material Constraints +11.3 points
• GPT-o3: Main Leaderboard down 14.2 points; Material Constraints -27.2 points
• Claude Sonnet 4.6: Main Leaderboard down 11.7 points; Code Execution -5 points; Material Constraints -19.9 points

Signals to Watch

• Claude Opus 4.7: Main Leaderboard sharp drop of 11.4 points
• Claude Sonnet 4.6: Main Leaderboard sharp drop of 11.7 points
• Grok 4: Material Constraints sharp drop of 16.5 points
• GPT-o3: Main Leaderboard sharp drop of 14.2 points
• GPT-6 Astra: Main Leaderboard sharp drop of 27.3 points
• Gemini 2.5 Pro: Material Constraints sharp drop of 16.5 points
• Gemini 3.1 Pro: Code Execution sharp drop of 30.5 points
• GLM-4.6: Data incomplete (missing judgment, integrity, and communication dimensions; API failure/timeout), entered automatic rerun, not ranked this period
• Qwen3 Max: Data incomplete (missing several evaluation dimensions; API failure/timeout), entered automatic rerun, not ranked this period

When reading this kind of Smoke brief, the focus should be 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 an early signal of real degradation, requiring follow-up runs for confirmation.

Data source: YZ Index | Run #426 | View raw data