About Winzheng
In 1998, Winzheng started as a software tech forum. Over 28 years, it grew to 1.5 million registered users, 500,000 published software titles, and the trust of generations.
In 2025, we made a decision: apply the same meticulous spirit to AI model evaluation.
Not just writing a few reviews. We built an automated system — running 212 questions weekly across 11 models, with real code execution, citation verification, and scores averaged over 5 runs. What comes out is what gets published — no coordination, no adjustments, no favoritism.
Today, 9 万+ people receive our evaluation results by email every week.
Our Products
YZ Index
Who changed? Why? Which AI to use today?
212 questions randomly sampled, code executed in sandbox, citations verified against source.
Rankings based on rolling 5-run averages, not single results.
11 models, 6 dimensions, weekly automated reports.
Research Lab
The leaderboard tells you "who's #1." Research Lab tells you "why."
Model safety, edge deployment, performance teardowns — not restating others' papers, but conclusions we ran ourselves.
News
We don't dump every AI headline on you.
We only publish what changes capabilities, pricing, stability, or your model selection.
Verify first, publish second. Impact over hype.
Why Trust Us
Contact Us
Business / General:info@winzheng.com
Research Lab：lab@winzheng.com
Article Submission:Online Submission
Citation guidelines:When citing YZ Index data or research reports, please credit "Source: Winzheng (winzheng.com)" with a link to the original.
Global Picks content is a compiled Chinese version. Please cite both this site and the original source when republishing.
Media Citation
We welcome citation of YZ Index data and Research Lab reports.
Please credit "Source: Winzheng (winzheng.com)" with a link to the original.