Discover quality AI tools
Bring together the world’s leading artificial intelligence tools to help improve work efficiency
Articles published by AnthonyPerez
Chinese AI models dominate OpenRouter's latest weekly usage rankings, with all top five spots held by domestic firms. Xiaomi's MiMo-V2.5 leads at 10.5 trillion tokens, up 12% weekly. DeepSeek claims second and fifth places, while Tencent HuanYuan 3, open-sourced July 6, surges over 999% to third.
Chinese AI models dominate OpenRouter's latest weekly usage rankings, with all top five spots held by domestic firms. Xiaomi's MiMo-V2.5 leads at 10.5 trillion tokens, up 12% weekly. DeepSeek claims second and fifth places, while Tencent HuanYuan 3, open-sourced July 6, surges over 999% to third.
OpenAI expanded ChatGPT parental notifications to alert linked parents when minors are banned for violence threats or cyberbullying. Developed with Moonshot, the upgrade covers risks of harming others. It also added a learning mode and rest reminders for teens.
OpenAI expanded ChatGPT parental notifications to alert linked parents when minors are banned for violence threats or cyberbullying. Developed with Moonshot, the upgrade covers risks of harming others. It also added a learning mode and rest reminders for teens.
ByteDance and Alibaba have discontinued their intelligent agent features on Doubao and Tongyi Qianwen, effective today, aligning with China's new AI personification interaction service regulations. The rules target emotional companion AI like virtual partners, while tool-type AI such as Q&A and copywriting remains unaffected. Safety assessments and algorithm filings are now mandatory.
ByteDance and Alibaba have discontinued their intelligent agent features on Doubao and Tongyi Qianwen, effective today, aligning with China's new AI personification interaction service regulations. The rules target emotional companion AI like virtual partners, while tool-type AI such as Q&A and copywriting remains unaffected. Safety assessments and algorithm filings are now mandatory.
Xiaohongshu became the exclusive Chinese internet rights holder for the 2026 FIFA World Cup, offering all 104 matches free. The move aims to attract male users and diversify its content ecosystem amid slowing growth. However, deeper strategy defends against AI-driven search disruption, as the platform seeks to retain its decision-making entrance role against rising AI assistants.
Xiaohongshu became the exclusive Chinese internet rights holder for the 2026 FIFA World Cup, offering all 104 matches free. The move aims to attract male users and diversify its content ecosystem amid slowing growth. However, deeper strategy defends against AI-driven search disruption, as the platform seeks to retain its decision-making entrance role against rising AI assistants.
Google has updated its spam policy to combat AI poisoning, formally classifying manipulative Generative Engine Optimization GEO tactics as spam. The new guidelines target websites that fabricate content to trick AI systems into indexing false information. Penalties include search ranking demotions, removal from AI Overview results, and potential site-wide bans from Google's index. Industry experts note the challenge remains complex due to advanced evasion techniques and the inherent difficulty in distinguishing malicious content from legitimate commercial material.
Google has updated its spam policy to combat AI poisoning, formally classifying manipulative Generative Engine Optimization GEO tactics as spam. The new guidelines target websites that fabricate content to trick AI systems into indexing false information. Penalties include search ranking demotions, removal from AI Overview results, and potential site-wide bans from Google's index. Industry experts note the challenge remains complex due to advanced evasion techniques and the inherent difficulty in distinguishing malicious content from legitimate commercial material.
Vbot Weita Power, a leader in embodied intelligence, secured nearly 500 million yuan in a record pre-A round led by Dongfang Jiafu, Huatai Zijin, and Fosun Ruizheng. Its first product, the "Super Dog" robot, has begun mass delivery after receiving pre-orders worth nearly 100 million yuan. The company is expanding production and transitioning research from quadruped robots to humanoid models, focusing on integrated AI systems to build a physical-world AI ecosystem.
Vbot Weita Power, a leader in embodied intelligence, secured nearly 500 million yuan in a record pre-A round led by Dongfang Jiafu, Huatai Zijin, and Fosun Ruizheng. Its first product, the "Super Dog" robot, has begun mass delivery after receiving pre-orders worth nearly 100 million yuan. The company is expanding production and transitioning research from quadruped robots to humanoid models, focusing on integrated AI systems to build a physical-world AI ecosystem.
Google DeepMind introduces TIPSv2 to address a key weakness in AI vision models poor localization of fine details despite strong global image understanding The new method enhances zero shot segmentation by 141 points via full area supervision reduces training parameters by 42 and uses multi granularity text descriptions The model sets new benchmarks in segmentation and outperforms larger models in retrieval and classification Code and weights are open sourced
Google DeepMind introduces TIPSv2 to address a key weakness in AI vision models poor localization of fine details despite strong global image understanding The new method enhances zero shot segmentation by 141 points via full area supervision reduces training parameters by 42 and uses multi granularity text descriptions The model sets new benchmarks in segmentation and outperforms larger models in retrieval and classification Code and weights are open sourced
SuperCLUE 2025 Chinese Large Model Report shows Claude-Opus-4.5-Reasoning leads with 68.25 pts, followed by Gemini-3-Pro-Preview and GPT-5.2. Domestic models Kimi-K2.5-Thinking and Qwen3-Max-Thinking placed 4th and 6th, excelling in code generation and mathematical reasoning. Overseas models remain ahead but Chinese open-source models show strong competitive potential.
SuperCLUE 2025 Chinese Large Model Report shows Claude-Opus-4.5-Reasoning leads with 68.25 pts, followed by Gemini-3-Pro-Preview and GPT-5.2. Domestic models Kimi-K2.5-Thinking and Qwen3-Max-Thinking placed 4th and 6th, excelling in code generation and mathematical reasoning. Overseas models remain ahead but Chinese open-source models show strong competitive potential.
