Doubao Pro Leads with 93.01 Points: 2026-11-20 Smoke Quick-Test Data Brief

The 2026-11-20 Winzheng YZ Index Smoke quick test covered 14 models, and Doubao Pro ranked first for the day with 93.01 points. Smoke is a daily 10-question quick test, suited to observing short-term signals; it is not equivalent to the conclusions of the weekly Full leaderboard.

This Smoke evaluation covered only two main-board dimensions: code execution and material constraints. The main-board formula is 0.55 × code execution + 0.45 × material constraints. Because the daily sample is small, single-day scores are better treated as monitoring signals than as long-term verdicts on model capability.

Daily Rankings

RankModelMain BoardCode ExecutionMaterial ConstraintsIntegrity
#1Doubao Pro93.0198.586.3pass
#2Claude Opus 4.792.1694.589.3pass
#3GPT-6 Astra88.9396.879.3pass
#4Gemini 2.5 Pro80.097586.3pass
#5GPT-6 Sol77.2971.884pass
#6Grok 474.4966.784pass
#7GPT-6.1 Sol72.7971.874pass
#8GPT-5.572.3971.873.1pass
#9Claude Sonnet 4.671.8970.973.1pass
#10GLM-4.664.694786.3pass
#11DeepSeek V4 Pro59.154774pass
#12GPT-o357.9245.573.1pass
#13GPT-6 Luna55.9444.669.8pass
#14Gemini 3.1 Pro55.0629.586.3pass

Reading the Data

In today's Smoke quick test, Doubao Pro combined 98.5 on code execution with 86.3 on material constraints for a main-board score of 93.01, while Claude Opus 4.7 achieved 92.16 on the main board through a more balanced 89.3 on material constraints. GPT-6 Astra scored 96.8 on code execution but 79.3 on material constraints, for 88.93 on the main board. Gemini 2.5 Pro paired 86.3 on material constraints with 75 on code execution for 80.09 on the main board, and GPT-6 Sol posted 71.8 on code execution and 84 on material constraints for 77.29. The leading models show differing emphases in how strength on code execution and material constraints are combined.

Gemini 2.5 Pro rose 42.3 points on the main board and 75 points on code execution; GPT-6 Astra rose 24.1 points on the main board and 47.5 points on code execution; Grok 4 rose 24.1 points on the main board and 43.2 points on code execution; Doubao Pro rose 21.2 points on the main board and 49.7 points on code execution while falling 13.7 points on material constraints. GPT-o3 fell 20.2 points on the main board, 28.3 points on code execution, and 10.2 points on material constraints. These shifts may stem from sampling variation in the questions and require follow-up runs for confirmation.

On anomalous signals, Claude Sonnet 4.6 saw material constraints plunge 15.2 points, GPT-o3 saw its main-board score plunge 20.2 points, GPT-6 Luna saw material constraints plunge 25.2 points, and Gemini 3.1 Pro saw its main-board score plunge 13.7 points. Qwen3 Max did not take part in the ranking because of incomplete data. Smoke is a small-sample, single-day signal, so the observations above should be treated with caution.

Key Changes

  • Gemini 2.5 Pro: main board up 42.3 points, code execution +75 points
  • GPT-6 Astra: main board up 24.1 points, code execution +47.5 points
  • Grok 4: main board up 24.1 points, code execution +43.2 points
  • Doubao Pro: main board up 21.2 points, code execution +49.7 points, material constraints -13.7 points
  • GPT-o3: main board down 20.2 points, code execution -28.3 points, material constraints -10.2 points

Signals to Watch

  • Claude Sonnet 4.6: material constraints plunge -15.2 points
  • GPT-o3: main board plunge -20.2 points
  • GPT-6 Luna: material constraints plunge -25.2 points
  • Gemini 3.1 Pro: main board plunge -13.7 points
  • Qwen3 Max: incomplete data (missing dimensions including execution, source verification, judgment, integrity, and communication; API failure/timeout); an automatic re-run has been scheduled and it does not take part in this period's ranking

When reading a Smoke brief like this, the focus should be on two questions: first, whether a given model shows the same type of weakness on several consecutive days; second, whether its integrity rating moves from pass to warn or fail. Large single-day swings in execution or constraint scores may come from question sampling, or they may be an early signal of genuine degradation, and they require follow-up runs for confirmation.


Data source: YZ Index | Run #436 | View raw data

This article is from Winzheng Index blog, translated in full by Winzheng (winzheng.com). Click here to view the original When republishing the translation, please credit the source. Thank you!