Many AI rankings are unreliable due to self-evaluation, fake code tests, single-run rankings, and sponsor influence. YZ Index from Winzheng disrupts this with rigorous methods like sandboxed execution, rolling averages, and zero-AI judging.
The YZ Index WDCD Commitment Test, launched by Winzheng (winzheng.com), uses a 3-round, 30-question design to precisely dissect AI’s “credibility crisis.” It exposes the hidden danger of AI failing to honor its promises, urging enterprises to move beyond flashy benchmark scores and focus on true reliability.
The YZ Index v7 introduces DCD (Dynamic Context Decay), a new experimental dimension that tests whether AI models can maintain hard constraints across multi-turn dialogues, addressing a critical gap in existing evaluations that only assess single-turn responses.
On May 1, 2026, YZ Index completed its largest evaluation roster update since launch last year, replacing 9 models and introducing 7 new flagships in a single sweep. This generational overhaul reflects the rapid pace of AI industry updates, where the evaluation system now needs to keep up with monthly rather than yearly iterations.
DeepSeek V3 exhibited a contradictory performance in this week's evaluation: significant improvements in multiple capability metrics, with the overall score rising from 52.9 to 66.6, but a cliff-like drop in the stability dimension. This phenomenon of "enhanced capabilities but unstable output" deserves in-depth analysis.
In this week's Winzheng AI evaluation, Doubao Pro's overall score increased by 16.1 points, but its stability dimension dropped sharply by 19.8 points to 34.7, revealing severe challenges in maintaining answer consistency. This phenomenon may result from technical adjustments like temperature parameter changes or model routing updates, reflecting a trade-off between capability enhancement and output predictability.
This week's YZ Index evaluation captures a rare synchronous improvement in the "task expression" dimension across 10 out of 11 mainstream AI models, while Claude Opus 4.6 uniquely breaks through in the "material constraint" dimension. The report analyzes these developments and offers developer selection advice for different application scenarios.
Grok 3's stability score crashed from 54.2 to 31.7 points in the latest Winzheng evaluation, exposing a fatal weakness in current AI models that excel at coding but fail at real-world engineering judgment.
GPT-o3's availability score plummeted from 100 to 69 in just one week, exposing fundamental architectural defects rather than isolated issues—a technical accident that reveals systemic imbalances in AI development.
Doubao Pro's stability score plummeted from 54.5 to 34.7 (a 36.3% drop) this week, despite significant improvements in programming and knowledge work dimensions, revealing a concerning pattern of "progress and regression coexisting" that warrants in-depth analysis.
Qwen Max exhibits extreme duality in this week's evaluation, with significant improvements in programming and long-context tasks, but a catastrophic decline in stability metrics. This "fire and ice" performance warrants in-depth analysis.
DeepSeek R1 shows extreme performance polarization in this week's evaluation: programming capability soared 47.4 points while stability plummeted 22.1 points, revealing critical trade-offs in model optimization.
Claude Opus 4.6's stability score crashed from 53.5 to 31.0 points this week, a 42.1% decline, while programming capabilities surged 208%, highlighting the complex trade-offs in AI model optimization.
Qwen Max experienced a significant decline in knowledge work performance this week, dropping from 81.6 to 71.8 points, primarily due to severe deterioration in logical reasoning tasks, particularly in classic "who lied" puzzles where scores fell from 50 to 25 points.
This analysis examines how 8 AI models performed on an engineering judgment task, revealing distinct capability tiers in identifying the typical "single success, batch failure" concurrency problem pattern.
In this engineering judgment test, 8 AI models demonstrated significant differences in understanding depth when diagnosing why identical code produces different results for different inputs.
In a database deletion recovery engineering judgment test, 8 mainstream AI models showed significant differences in understanding and response strategies. The models split into two distinct camps: 5 models scored 40 points by providing comprehensive solutions, while 3 models scored 0 by only addressing partial aspects of the problem.
Eight leading AI models showed clear capability divisions when tested on a seemingly simple time zone conversion question, with 5 models performing perfectly while 3 made calculation errors.
In a seemingly simple logical reasoning test, 8 mainstream AI models demonstrated starkly different performances with only a 50% success rate, exposing significant disparities in current AI's logical reasoning capabilities.
This week's YZ Index evaluation reveals a rare collective decline in knowledge work capabilities across AI models, with 6 out of 8 mainstream models showing performance degradation. Claude Sonnet 4.6 emerges as the only model with positive growth.
