YZ Index — AI Model Benchmarks, News & Research
Editor's Pick
巨浪对撞
3 Models Translation Showdown: Week 22 Quality Evaluation, gpt-o3 Leads with 8.3 Points
This week, 237 translation tasks were completed by 3 models. A blind evaluation of 3 samples across multiple models found gpt-o3 to be the best overall, with an
2026-05-25 07:02
Claude Suddenly Displays Hypnotic Instruction: Multiple Users Advised to Go to Sleep, Alignment Concerns Lurk Behind Anthropic’s Silence
On May 24, numerous users on X platform reported that Anthropic’s Claude model e
DeepSeek Welds V4-Pro's 75% Discount Permanent: A High-Stakes Bet to Reshape Global AI API Pricing Logic
DeepSeek's permanent 75% discount on V4-Pro signals a fundamental shift from tem
Overall Top 5
Full Rankings →
#1
Grok 4 83.7
▲2.7
·
#2
Claude Opus 4.7 81.9
▲1.9
·
#3
豆包 Pro 81.6
·
#4
Claude Sonnet 4.6 81.2
▼1.8
·
#5
DeepSeek V4 Pro 81.1
▲4.8
·
#6
Qwen3 Max 80.8
▲1.8
·
#7
GPT-5.5 79.4
▲2.4
·
#8
GPT-o3 78.5
·
#9
文心一言 4.5 74.2
▲7.1
·
#10
Gemini 3.1 Pro 52.8
▼24.9
·
#11
Gemini 2.5 Pro 49.3
▼29.7
·
▲ 文心一言 4.5 +70.7 · ▼ DeepSeek V3 -75.1
·
#1
Grok 4 83.7
▲2.7
·
#2
Claude Opus 4.7 81.9
▲1.9
·
#3
豆包 Pro 81.6
·
#4
Claude Sonnet 4.6 81.2
▼1.8
·
#5
DeepSeek V4 Pro 81.1
▲4.8
·
#6
Qwen3 Max 80.8
▲1.8
·
#7
GPT-5.5 79.4
▲2.4
·
#8
GPT-o3 78.5
·
#9
文心一言 4.5 74.2
▲7.1
·
#10
Gemini 3.1 Pro 52.8
▼24.9
·
#11
Gemini 2.5 Pro 49.3
▼29.7
·
▲ 文心一言 4.5 +70.7 · ▼ DeepSeek V3 -75.1
·
最新资讯
View All News →LQA Agent Reaches 90% Agreement with Human Reviewers: Smartling Bets on AI to Reshape Enterprise Localization
Smartling, a localization software service provider, announced on May 19 what it calls its "largest-ever" update to AI t
Claude Suddenly Displays Hypnotic Instruction: Multiple Users Advised to Go to Sleep, Alignment Concerns Lurk Behind Anthropic’s Silence
On May 24, numerous users on X platform reported that Anthropic’s Claude model exhibited an anomalous behavior jokingly
DeepSeek Welds V4-Pro's 75% Discount Permanent: A High-Stakes Bet to Reshape Global AI API Pricing Logic
DeepSeek's permanent 75% discount on V4-Pro signals a fundamental shift from temporary promotion to permanent pricing, e
Taiwan Launches National AI Strategy Committee: Risk Assessment by July, Industry Regulations by 2028, Asia-Pacific Governance Race Quietly Accelerates
Taiwan has established a National AI Strategy Committee chaired by the Premier, initiating the implementation of the AI
MLCommons公布2026 Rising Stars：39位机器学习系统新星入选
MLCommons公布第四届 Rising Stars 名单，39位来自全球26所机构的早期研究者从175多名申请者中脱颖而出。入选者研究覆盖大语言模型、ML系统效率、软硬件协同设计、可信AI、多模态学习及医疗、网络安全、科学计算等应用方向
Modal Labs $355M Series C Funding: 5x ARR Growth Leads Serverless GPU
Modal Labs disclosed a $355 million Series C funding on May 21, 2026, reflecting strong market demand for serverless GPU
Cohere Open-sources Command A+ 218B MoE Model to Reshape Enterprise Sovereign AI
Cohere has open-sourced Command A+, a 218B-parameter sparse MoE model with 25B active parameters and 128K context length
US Revokes 90-Day Federal Review Order for Frontier AI Models, Highlighting Regulatory Divergence Among US, China, and EU
The US abruptly withdrew a planned 90-day federal review requirement for frontier AI models hours before its signing, ci
AI安全实时博弈：连谷歌也在摸着石头过河
我们正处在AI安全过渡期——所有人都一样。无论巨头还是初创，都在实时应对前所未有的挑战。谷歌的安全举措暴露了系统性难题：传统安全框架失效，攻防博弈加速，监管滞后。本文深度解析AI安全现状，探讨行业如何从“被动应急”走向“主动防御”。
ERNIE 4.5 Code Execution Plummets from 95 to 50, Main Score Drops 27.2 Points in a Single Day
In today's Smoke evaluation, ERNIE 4.5's main score fell from 88.48 to 61.25, a single-day drop of 27.2 points, driven b
DeepSeek V4 Pro Integrity Rating Switches from Fail to Pass, Main Ranking Surges 23 Points in a Single Day
DeepSeek V4 Pro's integrity rating on today's Smoke evaluation jumped directly from Fail to Pass, and its main ranking s
DeepSeek V4 Pro Tops with 97.08 Points, 文心一言 Execution Score Plunges 27.2 Points
In the latest Smoke Lightweight Benchmark, DeepSeek V4 Pro scored 97.08 to become the only model breaking 97, while 文心一言
深度横评
查看全部 →MLCommons公布2026 Rising Stars：39位机器学习系统新星入选
MLCommons公布第四届 Rising Stars 名单，39位来自全球26所机构的早期研究者从175多名申请者中脱颖而出。入选者研究覆盖大语言模型、ML系统效率、软硬件协同设计、可信AI、多模态学习及医疗、网络安全、科学计算等应用方向
ERNIE 4.5 Code Execution Plummets from 95 to 50, Main Score Drops 27.2 Points in a Single Day
In today's Smoke evaluation, ERNIE 4.5's main score fell from 88.48 to 61.25, a single-day drop of 27.2 points, driven b
DeepSeek V4 Pro Integrity Rating Switches from Fail to Pass, Main Ranking Surges 23 Points in a Single Day
DeepSeek V4 Pro's integrity rating on today's Smoke evaluation jumped directly from Fail to Pass, and its main ranking s
WDCD Compliance
#1
Claude Opus 4.7
65
#2
Claude Sonnet 4.6
62.5
#3
豆包 Pro
60
#4
Gemini 2.5 Pro
57.5
#5
Qwen3 Max
57.5
#6
GPT-o3
55
#7
文心一言 4.5
52.5
View full compliance rankings →
Research Lab
3 Models Translation Showdown: Week 22 Quality Evaluation, gpt-o3 Leads with 8.3 Points
This week, 237 translation tasks were completed by 3 models. A blind evaluation of 3 samples across
WDCD Run #125: Average Instruction Decay Hits 63.6%, Claude Opus 4.7 Leads with Only 30% Drop
WDCD Run #125 (2026-05-20) tested 11 large language models on multi-turn commitment integrity, with
3 Major Model Translation Showdown: Week 21 Quality Evaluation, gpt-o3 Leads with 8.7 Points
This week, 242 translation tasks were completed by 3 models. 3 articles were sampled for multi-model