GPT-5.5, GPT-6 Astra, GPT-6 Luna, and GPT-6 Sol Tie for First at 87.27: 2026-11-03 Smoke Quick-Test Data Brief

On 2026-11-03, the YZ Index Smoke quick test covered 13 models. GPT-5.5, GPT-6 Astra, GPT-6 Luna, and GPT-6 Sol tied for first place that day with 87.27 points. Smoke is a daily 10-question quick test, suitable for observing short-term signals; it is not equivalent to conclusions from the Full weekly leaderboard.

This Smoke evaluation covered 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 as long-term conclusions about model capabilities.

Daily Ranking

Data Interpretation

In today's YZ Index Smoke quick test, GPT-5.5, GPT-6 Astra, GPT-6 Luna, and GPT-6 Sol all scored 87.27 on the main leaderboard, with code execution at 100 and material constraints at 71.7 for all four, showing a structure that combines high code execution scores with mid-range material constraints. Claude Sonnet 4.6 scored 84.9 on the main leaderboard, with code execution at 88.9 and material constraints at 80, presenting a relatively balanced score distribution. DeepSeek V4 Pro and Claude Opus 4.7 ranked in the middle, with combinations of code execution 75 and material constraints 80, and code execution 70 and material constraints 83.3, respectively; their main leaderboard scores were 77.25 and 75.99.

Compared with the previous run using the same methodology, DeepSeek V4 Pro's main leaderboard score rose by 43.5 points, with code execution up 75 points and material constraints up 5 points; Gemini 3.1 Pro's main leaderboard score rose by 27.8 points, with code execution up 71.9 points and material constraints down 26.1 points; Doubao Pro's main leaderboard score rose by 23.4 points, with code execution up 63.9 points and material constraints down 26.1 points; GPT-o3's main leaderboard score rose by 15.8 points, with code execution up 44.5 points and material constraints down 19.2 points; GPT-5.5's main leaderboard score rose by 14.8 points, with code execution up 50 points and material constraints down 28.3 points. All of these increases and decreases come from single-day small-sample tests.

The above score changes may stem from question sampling fluctuations, or they may reflect real differences in model performance under specific constraints; follow-up runs are needed to verify stability. As a small-sample single-day signal, the current Smoke quick-test data is for observation on the day only and should not be used for long-term inference.

Main Changes

• DeepSeek V4 Pro: Main leaderboard up 43.5 points, code execution +75 points, material constraints +5 points
• Gemini 3.1 Pro: Main leaderboard up 27.8 points, code execution +71.9 points, material constraints -26.1 points
• Doubao Pro: Main leaderboard up 23.4 points, code execution +63.9 points, material constraints -26.1 points
• GPT-o3: Main leaderboard up 15.8 points, code execution +44.5 points, material constraints -19.2 points
• GPT-5.5: Main leaderboard up 14.8 points, code execution +50 points, material constraints -28.3 points

Signals to Watch

• No publishable anomalous signals were retained in this run.

When reading this type of Smoke brief, focus on two questions: first, whether a model exposes the same type of weakness for multiple consecutive days; second, whether its integrity rating moves from pass to warn or fail. Large single-day changes in execution or constraint scores may come from question sampling, or they may be early signals of real degradation, requiring follow-up runs for verification.

Data source: YZ Index | Run #409 | View raw data