GPT-6 Sol Tops with 90.51 Points: 2026-10-30 Smoke Quick-Test Data Brief

The 2026-10-30 YZ Index Smoke quick test covered 13 models, with GPT-6 Sol ranking first for the day at 90.51 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals and 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 capabilities.

Daily Ranking

Data Interpretation

Today, the top three on the main leaderboard all showed perfect scores in code execution. GPT-6 Sol ranked first with 90.51 on the main leaderboard, with 100 in code execution and 78.9 in material constraints; GPT-6 Astra and GPT-6.1 Sol both scored 87.4 on the main leaderboard, with 100 in code execution and 72 in material constraints. GPT-6 Luna scored 80.42 on the main leaderboard, with 74.3 in code execution and 87.9 in material constraints, forming a profile with relatively weaker code and stronger material constraints. GPT-5.5 scored 79.1 on the main leaderboard, with 75 in code execution and 84.1 in material constraints, contrasting with Gemini 3.1 Pro's 67.7 in code execution and 84.1 in material constraints.

DeepSeek V4 Pro fell 27.3 points on the main leaderboard and 50 points in code execution; Gemini 2.5 Pro fell 24.2 points on the main leaderboard and 46.5 points in code execution; Doubao Pro fell 23.6 points on the main leaderboard and 55.2 points in code execution while gaining 15 points in material constraints; GPT-o3 fell 16.2 points on the main leaderboard and 41.2 points in code execution while gaining 14.4 points in material constraints; GPT-6 Luna rose 13.3 points on the main leaderboard and 30.3 points in material constraints. Claude Sonnet 4.6 plunged 12.9 points on the main leaderboard, Claude Opus 4.7 plunged 17.4 points in code execution, and Grok 4 plunged 12.2 points on the main leaderboard. These single-day swings may stem from sampling variation in the questions or may reflect real degradation; follow-up runs are needed to confirm.

GLM-4.6 and Qwen3 Max had incomplete data due to API failures and were not included in this issue's ranking. As a small-sample, single-day signal, Smoke's observations above are for reference only and should not be used for long-term inference.

Key Changes

• DeepSeek V4 Pro: Main leaderboard down 27.3 points, code execution down 50 points, integrity pass→warn
• Gemini 2.5 Pro: Main leaderboard down 24.2 points, code execution down 46.5 points
• Doubao Pro: Main leaderboard down 23.6 points, code execution down 55.2 points, material constraints up 15 points
• GPT-o3: Main leaderboard down 16.2 points, code execution down 41.2 points, material constraints up 14.4 points
• GPT-6 Luna: Main leaderboard up 13.3 points, material constraints up 30.3 points

Signals to Watch

• Claude Sonnet 4.6: Main leaderboard plunged 12.9 points
• Claude Opus 4.7: Code execution plunged 17.4 points
• GPT-o3: Main leaderboard plunged 16.2 points
• Doubao Pro: Main leaderboard plunged 23.6 points
• Grok 4: Main leaderboard plunged 12.2 points
• Gemini 2.5 Pro: Main leaderboard plunged 24.2 points
• DeepSeek V4 Pro: Main leaderboard plunged 27.3 points
• GLM-4.6: Data incomplete (missing communication dimension; API failure/timeout), automatically queued for re-run, not included in this issue's ranking
• Qwen3 Max: Data incomplete (multiple evaluation dimensions missing; API failure/timeout), automatically queued for re-run, not included in this issue's ranking

When reading this kind of Smoke brief, focus on two questions: first, whether a model has shown 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, requiring follow-up runs to confirm.

Data source: YZ Index | Run #402 | View raw data