Welcome to the 【AI Daily】 section! This is your guide to exploring the world of artificial intelligence every day. Here we present the highlights of the AI field for developers, helping you gain insights into technological trends and learn about innovative AI product applications.
Fresh AI products click to learn more: https://top.aibase.com/
1. Qwen3-Embedding series models officially released by Qwen
I am very excited as an AI assistant that the Qwen team has launched the Qwen3-Embedding series models. This new model not only performs exceptionally well in multi-language text understanding and retrieval tasks but also offers flexible configuration options and strong multi-language support, showcasing its leading position in the field of text processing.
【AiBase Summary:】
📚 The Qwen3-Embedding series is based on the Qwen3 base model, providing three configurations with parameter sizes ranging from 0.6B to 8B, suitable for different performance and efficiency requirements in various scenarios.
🌍 Supports over 100 languages, featuring powerful multi-language, cross-language, and code retrieval capabilities, designed with dual-tower and single-tower structures.
🌟 Scores 70.58 on the MTEB multilingual leaderboard, outperforming many commercial API services, demonstrating excellent text representation and ranking capabilities.
Details link: https://modelscope.cn/collections/Qwen3-Embedding-3edc3762d50f48
2. ByteDance releases image editing model SeedEdit 3.0, with further enhanced detail retention capability
SeedEdit 3.0 is an image editing model developed based on Seedream 3.0. By using diversified data fusion and a specialized reward model, it significantly enhances the ability to retain the subject, process background details, and follow instructions, especially excelling in portrait editing, background replacement, and complex light and shadow handling.
【AiBase Summary:】
✨ Introduces efficient data fusion strategies and a specialized reward model, significantly enhancing image editing retention effects.
🌟 Supports 4K resolution editing, displaying strong detail processing capabilities when handling complex scenes such as portraits and light and shadow changes.
🚀 Inference accelerated to the 10-second level, leading in 23 categories of editing task evaluations, with usability increased to 56.1%.
Details link: https://seed.bytedance.com/seededit
3. The strongest AI voice on Earth! Eleven v3 Alpha version震撼ly released, capable of ‘acting’ as well as speaking
The Eleven v3 Alpha version released by ElevenLabs is a milestone in the TTS (Text-to-Speech) field due to its outstanding emotional expression, multi-language support, and natural dialogue capabilities, redefining text-to-speech technology.
【AiBase Summary:】
🌟 The Eleven v3 Alpha version introduces audio tags, enabling precise control over emotion, speed, and adding sound effects, making the voice more realistic and expressive.
🌐 Supports over 70 languages, with multi-role dialogue capabilities, applicable in various scenarios such as film dubbing, education, and customer service.
🚀 After technical upgrades, text understanding and dialogue generation capabilities have significantly improved, simplifying the creative process with automatic tagging features, enabling non-professionals to easily generate high-quality voice content.
4. Anthropic launches AI models customized for national security, supported by Amazon and Google
Anthropic released the Claude Gov model suite, specifically designed for national security agencies, enhancing the handling of confidential materials and receiving strategic support from Amazon and Google, but faces legal action from Reddit.
【AiBase Summary:】
🌐 The Claude Gov model suite is specifically designed for national security agencies, enhancing the handling of confidential materials.
🤝 The product receives support from Amazon and Google, available only to institutions with the highest security clearance.
⚖️ Anthropic faces legal action from Reddit, accusing it of unauthorized use of user data for model training.
5. Keeling AI monthly subscription revenue exceeds 100 million yuan for two consecutive months, with a user base exceeding 22 million
Keeling AI surpassed a $100 million annualized revenue run rate within 10 months, with P-end paid subscription members contributing most of the income, and the global user base surpassing 22 million.
【AiBase Summary:】
✨ Keeling AI's annualized revenue run rate exceeded $100 million within 10 months.
💰 P-end paid subscription members contribute nearly 70% of total revenue.
👥 The global user base exceeds 22 million, providing API services for enterprise customers.
6. Meta releases technical details of Aria Gen2: Four cameras, 8 hours of battery life challenge Apple Vision Pro
Meta fully disclosed the technical details of the Aria Gen2 research glasses for the first time. Compared to the first generation, it has achieved comprehensive upgrades in hardware design, sensor technology, and AI processing capabilities.
【AiBase Summary:】
Four cameras加持, global shutter sensors solve motion distortion issues, with significantly improved depth measurement accuracy.
New contact microphones, nasal bridge integrated structural sound conduction technology, clear audio pickup even in noisy environments.
AI processing capability greatly enhanced, supporting six degrees of freedom position tracking, eye movement tracking, and 3D hand tracking, laying the foundation for future AR interactions.
7. LovePoet Technology's PixVerse domestic version 'TakeMeAI' officially launched
LovePoet Technology's PixVerse domestic version 'TakeMeAI' has officially launched, supporting both web and mobile platforms, and providing an open API platform, significantly reducing video production costs and time.
【AiBase Summary:】
TakeMeAI helps users easily create personalized video content through AI effects and WoW launchers.
The domestic version supports V4.5, providing convenient video generation solutions to meet various needs.
TakeMeAI open platform collaborates with multiple top enterprises, offering efficient video generation tools for enterprise users.
Details link: https://pai.video
8. Wells Fargo boldly predicts: ChatGPT advertising revenue will reach $100 billion by 2030
Analysts at Wells Fargo predict that by 2030, ChatGPT will account for 30% of the global search advertising market, with annual revenue approaching $100 billion, challenging Google's dominant position.
【AiBase Summary:】
By 2030, ChatGPT is expected to account for 30% of the global search advertising market, with annual revenue approaching $100 billion.
Currently, Google dominates over 90% of the search advertising market, but it is expected to drop to around 60% by 2030.
ChatGPT's commercialization process may be driven by partnerships with mobile manufacturers and antitrust rulings.
9. Wang Ziru thanks Dong Mingzhu and Lei Jun, embarking on a second entrepreneurial journey as an AI review UP
Prominent tech blogger Wang Ziru announced his relaunch under the name 'Wang Ziru AI', focusing on AI content entrepreneurship and assisting traditional industries in digital transformation, while sharing experiences from his time at Gree and expressing gratitude to Dong Mingzhu and Lei Jun.
【AiBase Summary:】
🚀 On June 6, Wang Ziru's Bilibili account relaunched and was renamed 'Wang Ziru AI', starting a second entrepreneurial journey as an AI review UP.
💼 He once reshaped the sales system at Gree, thanks to encouragement from Dong Mingzhu and Lei Jun, continuing to pursue ideals.
💡 Choosing AI domain entrepreneurship because he sees its great potential, believing it can quickly yield returns.
10. Zhiyuan releases RoboOS2.0 and RoboBrain2.0: First robot supporting MCP mechanism
At the Beijing Zhiyuan Conference, Beijing Zhiyuan Artificial Intelligence Research Institute released the embodied intelligent operating system RoboOS2.0 and the large model RoboBrain2.0, promoting the development of the embodied intelligence ecosystem through open source initiatives.
【AiBase Summary:】
First robot operating system RoboOS2.0 supporting MCP mechanism, lowering development thresholds and improving multi-robot collaboration capabilities.
RoboBrain2.0 improves task planning accuracy by 74%, performing excellently in spatial reasoning and intelligent scheduling.
Already collaborating with multiple enterprises to jointly build an open and collaborative intelligent robotics ecosystem.
11. Google's blockbuster new product! Portraits allow you to interact with virtual experts, unlocking secrets of communication and leadership
Google's Portraits is an innovative product based on AI technology, allowing users to interact with virtual experts in real-time to learn communication and leadership skills, featuring high personalization and interactivity.
【AiBase Summary:】
🌟 Immersive dialogue learning experience, mastering practical skills by interacting with virtual experts.
🌐 AI-driven personalized learning, dynamically adjusting content to ensure relevance.
🌍 Wide range of application scenarios, from workplace to education, assisting personal and professional development.
12. OpenAudio releases open-source TTS model S1-Mini: 0.5B parameters create super-natural AI voice
Fish Audio released a lightweight version of the S1 model, S1-Mini, with only 0.5B parameters yet possessing high expressiveness and multi-language support. After being open-sourced, it significantly lowers development barriers, bringing innovative possibilities to the fields of education and entertainment.
【AiBase Summary:】
🌟 Lightweight design: 0.5B parameters, compatible with edge devices, supporting 14 languages and over 50 emotions.
🌐 Open-source empowerment: free download, lowering development barriers, promoting global technological popularization and innovation.
🚀 Outstanding performance: comparable to industry giants, particularly excelling in multi-language and complex dialogue scenarios.
Details link: https://huggingface.co/fishaudio/openaudio-s1-mini
13. AI-driven local video editing tool Diffusion Studio Pro, dubbed “CapCut + Cursor” combination
The AI-driven video editing tool Diffusion Studio Pro made its official debut, attracting widespread attention with its powerful AI functions and localized design. It combines the advantages of CapCut and Cursor, offering multi-modal AI-enabled nonlinear editing experiences, while supporting free use, significantly lowering the entry threshold for creation.
【AiBase Summary:】
🌟 Multi-modal AI enables nonlinear editing, with built-in intelligent agent sidebars automating workflows, significantly improving creation efficiency.
🔒 Local-first design protects privacy, attracting independent creators and small teams with free unlimited tier mode.
🌍 Supports wide application scenarios, from short videos to professional productions, providing full-chain support from creativity to launch.
14. Zhiyuan Research Institute releases Emu3 and other 'Wuzhai' series large models
At the seventh 'Beijing Zhiyuan Conference', the Zhiyuan Research Institute released the 'Wuzhai' series of large models, including Emu3, Jianwei Brainμ, RoboOS2.0, RoboBrain2.0, and OpenComplex2, covering multimodal intelligent technologies and promoting the application of artificial intelligence.
【AiBase Summary:】
🚀 Emu3, as a native multimodal world model, integrates visual, auditory, and tactile data, enhancing the machine's understanding of the world.
🧠 Jianwei Brainμ combines neuroscience results, providing biological support for the development of machine intelligence.
🤖 RoboOS2.0 and RoboBrain2.0 promote the embodied intelligence collaboration framework, accelerating the progress of robotic technology.