Welcome to the "AI Daily" column! This is your guide to exploring the world of artificial intelligence every day. Each day, we present you with the latest content in the AI field, focusing on developers to help you understand technical trends and innovative AI product applications.
New AI products Click for more information:https://app.aibase.com/zh
1. Shengshu Technology Launches Vidu Q2, Bringing More Realistic AI Performances Through Subtle Expressions!
The Vidu Q2 model launched by Shengshu Technology has made significant breakthroughs in the field of image-to-video, especially excelling in generating subtle expressions, providing a more realistic and vivid visual experience for AI performances.
【AiBase Highlights:】
🎭 Vidu Q2 can accurately capture subtle changes in expressions, enhancing the naturalness and emotional expression of video generation.
🎥 Supports multiple video modes, including image-to-video, start-end frame video, and adjustable duration options, meeting diverse needs.
💡 Shengshu Technology is committed to promoting the development of creative industries through AI technology, bringing users a higher quality audio-visual creation experience.
2. Volcano Engine Launches Lumi Platform, Supporting Visual Model Lora Fine-tuning
The Lumi platform launched by Volcano Engine now supports Lora fine-tuning for visual models such as Doubao and Jiemeng, aiming to help enterprises efficiently customize unique visual styles to meet market demands.
【AiBase Highlights:】
🧠 The Lumi platform supports Lora fine-tuning for visual models, helping enterprises customize unique visual styles.
🚀 The platform provides end-to-end services from image generation to video generation, meeting professional AIGC needs in enterprise scenarios.
💡 The Lumi platform helps enterprises efficiently build customized AIGC production capabilities, enhancing user experience.
3. Alibaba Cloud CTO Reveals: Tongyi Qianwen Has Open-Sourced Over 300 Models, With Download Volume Exceeding 600 Million
At the 2025 Yunqi Conference, Alibaba Cloud showcased significant achievements of the Tongyi Qianwen project, including open-sourcing over 300 models and exceeding 600 million downloads. This demonstrates Alibaba Cloud's strong influence and technical strength in the AI field, and it promotes technological innovation and application through open-source models.
【AiBase Highlights:】
🚀 The Tongyi Qianwen project has open-sourced over 300 models, demonstrating strong technical capabilities.
📊 Total downloads have exceeded 600 million, reflecting high user recognition of Alibaba Cloud's AI technology.
🖼️ Tongyi Wanxiang has generated over 390 million images and more than 70 million videos, showcasing digital content generation capabilities.
4. Baidu Opensources Qianfan-VL, Kunlun Chip Powers Multimodal AI Breakthroughs
Baidu officially open-sourced its latest visual understanding model, Qianfan-VL, which includes three versions: 3B, 8B, and 70B, suitable for different scenarios. The model has strong multimodal capabilities, especially in OCR and education fields, and its training relies on Baidu's self-developed Kunlun P800 chip.
【AiBase Highlights:】
🧠 Qianfan-VL is a powerful multimodal large model that can process both image and text information simultaneously.
💡 The Kunlun P800 chip supports the model's training, with low power consumption and high efficiency, optimizing large-scale computing performance.
🚀 The Qianfan-VL series has been open-sourced on GitHub and Hugging Face for free use by developers.
Details: https://github.com/baidubce/Qianfan-VL
5. Microsoft Integrates Anthropic AI Models, Expanding Copilot Assistant Capabilities
Microsoft announced the integration of Anthropic's AI models into the Copilot assistant, marking a new step in its diversified strategy in the generative AI field. Although Microsoft continues to maintain close collaboration with OpenAI, it begins to incorporate Anthropic's technology to meet the needs of commercial customers. Enterprise users can now use Anthropic's models to build AI agents, and these models will run on Amazon and Google Cloud.
【AiBase Highlights:】
🤖 Microsoft integrates Anthropic's AI models into the Copilot assistant, promoting product diversification.
🔄 Although Microsoft maintains a close relationship with OpenAI, it is gradually adopting Anthropic's technology.
🚀 Enterprise users can choose Anthropic models to build AI agents, which require administrator activation before use.
6. OpenAI Builds Five New Data Centers in the US, Accelerating the Stargate Project
OpenAI announced the construction of five new data centers in the United States to enhance the computational capacity of the Stargate project. This project, initiated by multiple companies with a total investment of $50 billion, aims to promote the development of generative AI.
【AiBase Highlights:】
🌐 OpenAI will build five new data centers in the US, with the Stargate project's total computing capacity reaching nearly 7GW.
💼 Oracle will be responsible for constructing three new data centers, with the Abilene data center expanding and adding 600MW of computing capacity.
🚀 OpenAI plans to add 1GW of AI infrastructure each week in the future to drive further development of AI technology.
7. NVIDIA Open Sources Audio2Face Model, AI Helps Generate Real-Time Facial Animations
NVIDIA open-sourced its generative AI facial animation model, Audio2Face, providing SDKs and training frameworks that support offline and real-time processing, applicable to games, films, and other fields. This technology has been adopted by multiple game developers, enhancing the realism and immersive experience of characters.
【AiBase Highlights:】
🔊 NVIDIA open-sources the Audio2Face model, improving virtual character facial animation generation technology.
🎮 Supports offline rendering and real-time streaming processing, applicable to various scenarios.
🌟 Has been adopted by multiple game developers, simplifying the production process and enhancing character realism.
Details: https://build.nvidia.com/nvidia/audio2face-3d
8. Meta Releases Code World Model CWM: A 32B AI with Sandbox Simulation Capabilities
Meta's Code World Model (CWM) is an AI system with 32B parameters that can simulate and reason about code through a sandbox environment, reducing errors and improving debugging efficiency. The model requires high hardware specifications, needing dual H100 GPUs and RDMA technology support.
【AiBase Highlights:】
🧠 CWM simulates code in a sandbox environment before generating it, predicting the outcomes of code execution.
🔍 It quickly identifies code errors, improving debugging efficiency.
🚨 It can warn about potential risks before executing commands, enhancing security.
Details: https://github.com/facebookresearch/cwm
