Gemini 3.1 Pro Leads with 90.36: 2026-10-31 Smoke Rapid Test Data Brief

The 2026-10-31 YZ Index Smoke rapid test covers 14 models, and Gemini 3.1 Pro ranked first for the day with 90.36 points. Smoke is a daily 10-question rapid test, suitable for observing short-term signals and not equivalent to conclusions from the weekly Full 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 today's YZ Index Smoke rapid test, Gemini 3.1 Pro ranked first with a main leaderboard score of 90.36; its combination of 94 in code execution and 85.9 in material constraints shows relatively balanced performance. Doubao Pro had a main leaderboard score of 88.48, with code execution at 81.5 and material constraints at 97, showing a stronger material-constraints profile. Leading models such as GPT-6 Astra (main leaderboard 86.44, code execution 94, material constraints 77.2) and Claude Sonnet 4.6 (main leaderboard 85.18, code execution 94, material constraints 74.4) generally relied on a high code-execution level of 94 to support their overall scores.

DeepSeek V4 Pro gained 38.5 points on the main leaderboard and 69 points in code execution; Gemini 2.5 Pro gained 31 points on the main leaderboard and 57 points in code execution; Doubao Pro gained 27 points on the main leaderboard, 36.7 points in code execution, and 15.1 points in material constraints; Grok 4 gained 24.8 points on the main leaderboard and 45.9 points in code execution. These models' significant gains mainly came from improvements in code execution; GPT-6 Sol fell 16.4 points on the main leaderboard and 33.2 points in code execution, a clear decline.

Anomalous signals such as GPT-6 Luna's material constraints dropping sharply by 15.9 points, GPT-6 Sol's main leaderboard score dropping sharply by 16.4 points, GPT-5.5's main leaderboard score dropping sharply by 8.8 points, and GLM-4.6's material constraints dropping sharply by 43.6 points may stem from question sampling fluctuations or real degradation and require follow-up runs for confirmation; Qwen3 Max was not ranked due to incomplete data. As a small-sample single-day signal, Smoke should be interpreted with restraint.

Key Changes

• DeepSeek V4 Pro: Main leaderboard up 38.5 points, code execution +69 points, integrity warn→pass
• Gemini 2.5 Pro: Main leaderboard up 31 points, code execution +57 points
• Doubao Pro: Main leaderboard up 27 points, code execution +36.7 points, material constraints +15.1 points
• Grok 4: Main leaderboard up 24.8 points, code execution +45.9 points
• GPT-6 Sol: Main leaderboard down 16.4 points, code execution -33.2 points

Signals to Watch

• GPT-6 Luna: Material constraints dropped sharply by 15.9 points
• GPT-6 Sol: Main leaderboard dropped sharply by 16.4 points
• GPT-5.5: Main leaderboard dropped sharply by 8.8 points
• GLM-4.6: Material constraints dropped sharply by 43.6 points
• Qwen3 Max: Incomplete data (missing several evaluation dimensions; API failure/timeout), automatically queued for rerun, not ranked in this period

When reading this type of Smoke brief, focus on two questions: first, whether a model has exposed the same type of weakness for multiple consecutive days; second, whether its integrity rating has moved from pass to warn or fail. Large single-day changes in code-execution or material-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 #404 | View raw data