Welcome to the 【AI Daily】 section! Here is your guide to exploring the world of artificial intelligence every day. We present you with hot content from the AI field daily, focusing on developers and helping you gain insights into technical trends and innovative AI product applications.
Fresh AI products click to learn more: https://top.aibase.com/
1. No invitation code required! The domestic AI agent Manus opens registration for everyone.
The domestic AI agent platform Manus has recently opened user registration, allowing users to register directly without waiting and earn points as a reward. This platform, launched by Monica Company, quickly gained popularity due to the unique charm of its general-purpose AI Agent. Manus enhances user creativity through innovative engineering architecture that combines thinking and action, covering task execution in multiple fields, marking a shift in AI application.
【AiBase Summary:】
🚀 Manus platform is now open for registration, allowing users to execute tasks daily for free and earn point rewards.
🤖 Manus combines engineering innovation to integrate thinking and action, supporting the execution of complex tasks.
🌐 Collaborating with the Alibaba Tongyi Qwen team to explore the application of domestic models and computing platforms, enhancing user experience.
2. Global first design Agent Lovart enters beta testing!
Global first design Agent Lovart AI has launched, integrating image, video, and music models, revolutionizing traditional design workflows. Users can achieve full-chain automated design through natural language instructions, attracting attention from designers and start-ups. Its core features include task decomposition, cross-media content creation, and intelligent graphic separation, significantly improving design efficiency. Lovart performed exceptionally well during the Beta test phase, expected to further drive the development of the AI design tools market.
【AiBase Summary:】
🛠️ Lovart AI achieves full-chain automated design through natural language instructions; users only need one sentence to complete complex tasks.
🎥 Supports multi-modal model integration, enabling users to easily generate image, video, and music content, enhancing creative flexibility.
🌐 Lovart received enthusiastic feedback during the Beta test phase; users praised its five-fold increase in design efficiency, showcasing strong market potential.
Details link: https://lovart.ai
3. Ministry of Education defines boundaries for AI use: stage-based guidance from primary to high school, prohibiting copying of AI answers.
The Ministry of Education's "General Guidelines for Artificial Intelligence Education in Primary and Secondary Schools (2025 Edition)" and "Guidelines for the Use of Generative AI by Primary and Secondary School Students (2025 Edition)" clearly define the norms for using artificial intelligence in education, emphasizing that students should not directly copy AI-generated content to prevent impacts on independent thinking abilities. At the same time, guidelines propose stage-based strategies based on the characteristics of different grades, strengthening teachers' guiding responsibilities to ensure students use AI tools safely and reasonably.
【AiBase Summary:】
🚫 Prohibits students from directly copying AI-generated content as homework or exam answers, preventing reliance on AI from weakening independent thinking skills.
👩🏫 Teachers must guide students to conduct critical thinking training, analyzing logical flaws and cultural biases in AI-generated texts.
🔒 Establishes an AI tool whitelist system to ensure AI tools used in schools meet educational needs and data security compliance.
4. Building a Controllable Interactive World from Images! Kunlun Weiwei officially opens-source Matrix-Game.
Kunlun Weiwei Group announced on May 13th the opening of the Matrix-Game large model, a leading model focusing on interactive world generation, marking a new breakthrough in this field. Not only is Matrix-Game the first industrial-level open-source 10B+ space intelligence large model, but it also achieves high-quality generation and precise control through its three core parts.
【AiBase Summary:】
🛠️ Matrix-Game is the first open-source 10B+ space intelligence large model, focusing on interactive world generation, setting a new technical limit.
📊 The model quantifies performance across four dimensions—visual quality, temporal quality, etc.—through the GameWorld Score evaluation system, surpassing existing baseline models.
🌍 Matrix-Game has multi-scenario generalization capabilities, supporting different terrains and environments, enhancing user interaction experiences.
Details link: https://matrix-game-homepage.github.io
5. Google Gemma AI Model Downloads Exceed 150 Million Times
Google recently announced that downloads of its Gemma AI model collection have exceeded 150 million times, highlighting significant progress in the AI field. Gemma is a multimodal model capable of processing text and images, supporting over 100 languages, meeting global user needs. Despite the impressive download volume, it still falls short of its competitor Llama’s 1.2 billion downloads and has faced criticism from developers due to complex license terms.
【AiBase Summary:】
🌟 Downloads exceed 150 million times: Google’s Gemma AI model collection has attracted significant developer attention, marking significant progress.
🖼️ Multimodal functionality: Gemma supports text and image processing, meeting global user needs and supporting over 100 languages.
⚠️ License term controversy: Both Gemma and Llama face criticism due to complex license terms, with developers concerned about risks in commercial use.
6. OpenAI Launches One-click PDF Export Function for Deep Research Reports
OpenAI recently introduced a new feature allowing users to export their deep research reports as PDFs with one click. This update greatly enhances the practicality and sharing efficiency of reports, especially in enterprise environments. Users can easily generate professional documents containing citations, tables, and images, better supporting decision-making and communication.
【AiBase Summary:】
📝 The new PDF export function allows users to generate complete formatted deep research reports with one click, improving sharing efficiency.
🔗 This feature supports both newly generated and historical reports, ensuring users can access previous results anytime, enhancing report utility.
🚀 OpenAI advances the practical implementation of AI tools through this update, seamlessly integrating cutting-edge AI technology with existing enterprise workflows.
7. ByteDance Releases Unified Image Customization Framework DreamO: Integrates Dressing, Face Swap, and Style Migration
ByteDance launched DreamO on the Hugging Face platform, a brand-new framework integrating various image editing functions, marking a major breakthrough in AI image editing technology. DreamO not only supports dressing, face swapping, and style migration but also attracts developers through its open-source nature, driving industry innovation. Its one-stop solution lowers the usage threshold, making it easy for professional designers and ordinary users to achieve high-quality image editing.
【AiBase Summary:】
✨ DreamO is a unified framework integrating image dressing, face swapping, style migration, and other functions, supporting diverse image editing needs.
🔧 Through flexible parameter settings, users can easily perform precise image editing, lowering the usage threshold for all types of users.
🌍 The open-source nature of DreamO encourages participation from global developers, promoting innovation and application in the AI image editing field.
Details link: https://github.com/bytedance/DreamO
8. Moonlight Dark Side Kimi Interface Upgrade: Bubble Style Optimization, Changed to Boundary-free Content Display
Moonlight Dark Side Technology Co., Ltd.'s smart assistant Kimi underwent a major interface upgrade aimed at enhancing user experience and information presentation capabilities. This upgrade introduces multi-modal containers, supporting real-time K-line charts, HTML previews, and voice broadcasting functions, enhancing user interaction with Kimi. Although the new interface has sparked nostalgia for the old version among some users, the company emphasizes that this is part of Kimi's growth, hoping these updates will make it a stronger intelligent assistant.
【AiBase Summary:】
📈 Kimi supports real-time K-line charts, providing intuitive market trend analysis.
🎤 New voice broadcast function helps users track information.
🔧 Users can view Kimi service status in real-time, enhancing system transparency.
9. Alibaba MNN Major Update! Mobile Open-source Multi-modal AI Supports Qwen-2.5, Text, Image, Voice All-in-one!
Alibaba's MNN project released a new version of MnnLlmApp, supporting Qwen-2.5-Omni-3B and 7B models, greatly enhancing mobile multi-modal AI capabilities. This application supports multiple modal tasks including text, image, and audio generation and recognition, with efficient performance and low resource consumption, attracting developers' attention.
【AiBase Summary:】
🛠️ The new version of MnnLlmApp integrates Qwen-2.5-Omni-3B and 7B models, supporting multi-modal tasks including text, image, and audio, enhancing overall processing capabilities.
📱 The application runs entirely locally, ensuring data privacy without internet connection, and demonstrates outstanding performance on CPU inference, far exceeding similar products.
🌐 MNN's open-source nature and support for mainstream models make it an ideal platform for mobile multi-modal AI development, driving rapid industry advancement.
Details link: https://github.com/alibaba/MNN/blob/master/apps/Android/MnnLlmChat/README.md
10. ChatGPT Launches SharePoint Connector, Making Enterprise Data Analysis More Efficient!
ChatGPT recently launched a new feature that allows users to access company data via the SharePoint connector. This feature is currently in testing and is designed specifically for ChatGPT Plus, Pro, and Team users. Through this integration, ChatGPT can analyze and summarize content from multiple SharePoint sites and provide source references, helping enterprises achieve more efficient data management and analysis.
【AiBase Summary:】
✅ ChatGPT, through the SharePoint connector, supports company data analysis and summarization.
📊 This new feature is available for ChatGPT Plus, Pro, and Team users, helping integrate information across departments.
🔒 Data privacy is ensured; ChatGPT only accesses data the user has permission to view.
11. A New Open-source Visual Encoder OpenVision is Released: A Powerful Choice Beyond CLIP and SigLIP
OpenVision, a new series of visual encoders developed by the University of California, Santa Cruz, provides developers and enterprises with flexible options to enhance the efficiency of image processing and understanding. OpenVision not only supports multiple models and flexible deployment architectures but also performs excellently in multi-modal benchmark tests.
【AiBase Summary:】
🔍 OpenVision offers 26 different models with parameter scales ranging from 5.9 million to 632 million, suitable for various applications.
⚙️ Flexible deployment architectures support server-level loads and edge computing, with models adapting patch sizes to optimize performance.
📊 Performs excellently in multi-modal benchmark tests, adopting progressive training strategies to improve training efficiency and ensure downstream performance.
Details link: https://ucsc-vlaa.github.io/OpenVision/
12. AI Breakthrough! Alibaba Cloud and DeepSeek Stand Out, Forbes Publishes China's Top 50 AI Enterprises List
The 2025 Forbes list of China's top 50 artificial intelligence technology enterprises highlights China's rise in the global AI arena. With the rapid expansion of the market scale, more than 4,500 AI companies have achieved scaled applications in intelligent manufacturing, healthcare, autonomous driving, and other fields. The rise of enterprises in central and western regions shows technological innovation breakthroughs. Eleven companies dominate the large model field, driving the commercialization of generative AI technology.
【AiBase Summary:】
🚀 There are more than 4,500 AI enterprises in China, covering multiple fields and driving industry progress.
🌍 The number of enterprises selected from central and western regions has significantly increased, showing technological innovation breakthroughs.
💡 Eleven enterprises dominate 22% of the market share in the large model field, promoting the commercialization of generative AI technology.