Skip to main content
YZ Index

Winzheng · Task Communication Rankings

Structured output quality and formatting compliance.

Overall Rankings Code Execution Rankings Grounding Rankings Engineering Judgment Rankings Task Communication Rankings Value Rankings Changes
Side Dimension Rankings — Communication + Judgment sub-dimension performance
# Model Task Communication Code Execution Overall
🥇 Doubao Pro doubao
93.3
86.6 76.7
🥈 GPT-5.5 gpt
90.8
87.6 80.5
🥉 GPT-6 Luna gpt
88.8
81.7 79.4
4 GPT-6 Astra gpt
86.1
90.5 80.4
5 GPT-6 Sol gpt
86.1
92.1 82.2
6 GPT-6.1 Sol gpt
86.1
90.5 78.6
7 Grok 4 grok
85.8
85.3 80.4
8 GPT-o3 gpt
78.3
84.4 80.6
9 Gemini 3.1 Pro gemini
75.8
73.8 74.2
10 Claude Sonnet 4.6 claude
75.4
86.6 80.1
11 Claude Opus 4.7 claude
72.8
81.5 81.5
12 DeepSeek V4 Pro DeepSeek
66.1
75.1 76.4
13 Gemini 2.5 Pro gemini
65.8
75.2 72.2
14 Qwen3 Max qwen
50.3
76.5 73.6
15 GLM-4.6 zhipu
25
49.5 58.8

Run #362 · Formula v7 · Judge v6.4 · Benchmark v7

Full evaluation: 128 questions from pool of 100yz.rankings.footer.full_eval_suffix · Latest Evaluation2026-10-05 05:15 SGT · Models15models

Winzheng YZ Index — Winzheng · YZ Index