Domestic Large Models Dominate Overseas Rankings! Hunyuan Hy3 Preview Tops Global Model Usage Ranking

In the latest released OpenRouter global large model API call volume ranking, the performance of domestic large models has attracted attention. Tencent's newly launched Hy3preview model has successfully surpassed many international competitors with its high usage frequency, securing the top spot on the overall list. This data not only reflects developers' recognition of the model's performance but also marks the accelerating growth of the influence of domestic large models in the global AI ecosystem.

Comprehensive Performance: Excellent in Tool Usage and Programming Scenarios

In addition to leading in total call volume, Hy3preview also performs outstandingly in specific professional scenarios. According to scenario-specific statistical data, the model ranks first in the "Tool Use" dimension, demonstrating strong logical execution and external collaboration capabilities; while in the highly challenging "Coding" field, Hy3preview remains in second place. This comprehensive development trend has made it one of the most competitive AI tools for global developers today.

Industry Synergy: Domestic AI is Making Strong Advances

Notably, besides Tencent, other domestic large model manufacturers also released significant progress on the same day. Jiequ Star officially announced the release of its new image generation and editing model Step Image Edit2, which is now fully available on the open platform. With the dense release and application of these high-performance models, the large model industry is shifting from a "parameter competition" to an "application effectiveness competition," and the advantages of domestic AI in core areas such as image processing, complex programming, and automation tools are gradually expanding.

OpenAI predicts ChatGPT subscription users will reach 122 million

OpenAI launched a low-cost subscription service called ChatGPT Go at $8 per month, which is now available in 171 countries, aiming to reduce the barrier to AI usage. At the same time, it plans to test free and Go versions with ads in the United States, and has established principles to protect user privacy and ensure that ads are independent from model responses. It is predicted that consumer subscription users will reach 122 million this year.

Cybersecurity authorities have lawfully investigated platforms such as Xianying and Jiemeng AI for their non-compliant AI content labeling practices

On April 28, the Cyberspace Administration of China lawfully investigated the apps "Xianying" and "Maoxiang" as well as the website "Jiemeng AI", as these platforms failed to clearly label AI-generated content in accordance with regulations, violating the Cybersecurity Law and the Interim Measures for the Management of Generative AI Services, due to inadequate implementation of the labeling requirements.

Google Gemini's proactive assistant rebranded as Daily Briefing, deeply integrated with multiple data sources for smart delivery

Google's AI platform Gemini plans to rename its 'Your Day' feature to Daily Briefing, marking the maturation of its proactive AI assistant strategy. This feature actively extracts and delivers personalized information related to the user's daily schedule by integrating search history, emails, and Gemini chat data.

Adobe Explosive Update: Firefly AI Agent Public Test, Photoshop Can Now Seamlessly Arrange?

Adobe launched the public test of Firefly AI Assistant on April 28th, targeting Creative Cloud Pro and Firefly Pro/Premium paying users. The core upgrade is evolving from a single generation tool to an intelligent agent across applications, allowing operation without mastering software like Photoshop or Premiere.

Google Tests YouTube Conversational AI Search: Supports Text Summarization and Multimodal Content Aggregation

Google is testing a 'Ask YouTube' generative AI search feature on YouTube, upgrading keyword retrieval to a conversational mode. This experiment targets U.S. users aged 18 and above with YouTube Premium, aiming to transform video search from content matching to deep information understanding and aggregation.