Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers, helping you understand technology trends and innovative AI product applications.
Hot AI products Click to learn more:https://app.aibase.com/zh
1、DeepSeek-V4 Preview Version Released: 1M Long Context Enters a Universal Era
The DeepSeek-V4 preview version has been officially released, featuring 1M long context capabilities and improved reasoning performance and application efficiency through structural innovation and Agent ecosystem adaptation.
【AiBase Summary:】
🧠 The DeepSeek-V4 preview version is now available and open-sourced, achieving standardized 1M long context capabilities.
⚙️ Two versions are available: Pro (performance close to top closed-source models) and Flash (cost-effective).
💡 DSA mechanism optimizes long context processing, reducing computational costs.
More details: https://huggingface.co/collections/deepseek-ai/deepseek-v4 https://modelscope.cn/collections/deepseek-ai/DeepSeek-V4
2、Tesla Announces In-Car Voice Access to Doubao
Tesla announced that its in-car voice service will integrate ByteDance's Doubao large model, marking an important update for the company in the Chinese market. This collaboration, provided by DeepSeekChat, adopts a dual-model deployment plan, further promoting the localization of in-car AI technology.
【AiBase Summary:】
🚗 Tesla's collaboration with ByteDance's Doubao large model marks a major update in in-car AI.
🚀 This integration will accelerate the commercialization of domestic AI technology in the in-car field.
🤝 Industry autonomous car manufacturers, tech giants, and AI startups are jointly driving the development of in-car AI technology.
3、Huawei Launches HarmonySpace 6: Smart Assistant Xiaoyi Upgraded, Support Chatting While Driving!
Huawei officially launched the new generation of HarmonySpace 6, bringing comprehensive upgrades in cabin perception, smart interaction, and mobile entertainment. The newly released cabin intelligent agent Xiaoyi uses the latest MoLA 2.0 architecture, featuring a multi-modal large model with billions of parameters, further enhancing the smart assistant's understanding, decision-making, and execution capabilities. The new Xiaoyi will become a scenario-based AI assistant that can perform navigation, vehicle control, chatting, and life services while driving, allowing users to communicate while driving and switch tasks at any time. In terms of perception and safety, Huawei introduced the first in-cabin AI multi-modal perception system AMS. The system consists of cameras, infrared cameras, and high-precision Starlight sensors, enabling full-range action perception and vital sign monitoring. Even in emergencies, the system can detect passengers' vital signs and provide accurate information for rescue. In terms of audiovisual experience, HarmonySpace 6 introduced the industry's first in-car dual-screen display, using two 17.2-inch 3.4K screens and supporting smart split-screen technology, allowing the co-driver to enjoy a wider viewing experience. At the same time, HUAWEI SOUND's AI tuning magic cube and no-microphone K歌 3.0 function add many highlights to the in-car entertainment experience. In addition, Huawei also launched a million-pixel color intelligent car light module that supports full-color projection and intelligent adjustment of light color temperature to adapt to different weather conditions, further enhancing safety. Additionally, the newly launched in-car laser projector HUAWEI XSCENE LCoS will provide users with a cinema-level viewing experience. To increase fun, Huawei also launched the AI companion robot "Hamon", which not only can stick to the in-car screen but also engage in dialogue with users, adding fun to travel.
【AiBase Summary:】
🌟 The new HarmonySpace 6 in-cabin system is launched, equipped with the Xiaoyi smart assistant, featuring a large model with billions of parameters.
🛡️ The in-cabin AI multi-modal perception system AMS realizes all-round action perception and vital sign monitoring, ensuring passenger safety.
🎥 The industry's first dual 17.2-inch 3.4K dual-screen and AI tuning magic cube enhance the in-car audiovisual experience.
4、Xiaomi Launches Full-Chain Voice Large Model MiMo-V2.5, TTS Can "Generate New Voice Style in One Sentence", ASR Open Source Supports Dialects and Multilingual Mix
Xiaomi launched the full-chain voice large model MiMo-V2.5, where TTS can "generate a new voice style in one sentence" and ASR supports dialects and multilingual mix with open source. This model series comprehensively covers the needs of voice input and output in the Agent era, making sound a programmable, creative, and replicable intelligent medium.
【AiBase Summary:】
🎙️ MiMo-V2.5-TTS series realizes the "language as control" voice generation paradigm, supporting fine control of speech rate, emotion, tone, etc., through natural language instructions.
🎧 MiMo-V2.5-ASR open-source model supports multiple Chinese dialects and bilingual mixed scenarios, with smooth and accurate transcription of professional terms and lyrics.
🚀 Three TTS models are temporarily free to open, and the MiMo-V2.5-ASR model weights and code are fully open-sourced, accelerating the implementation of the Agent ecosystem.
5、Soul Opens Real-Time Digital Human Generation Model SoulXFlashTalk with Sub-Second Latency
Soul AI Lab officially opened the industry's first real-time digital human generation model with 1.4 billion parameters, SoulXFlashTalk, which features sub-second latency and 32 frames per second high frame rate, providing the industry with a complete real-time interaction solution. This open source not only reduces the R&D threshold but also provides basic support for more innovative projects. Soul's open-source strategy aims to promote the common development of the AI ecosystem and plans to open more AI technology achievements in the future, promoting the integration of AI and social fields.
【AiBase Summary:】
🌟 Soul AI Lab launched a 1.4B parameter real-time digital human generation model SoulXFlashTalk, with fast response capability and high frame rate.
🚀 Open source brings project pages, technical reports, source code, and model weights, lowering the R&D threshold and encouraging more innovation.
🤝 Soul will continue to promote the open-source strategy, drive the innovation of AI and social fields, and build a multimodal AI ecosystem.
6、Meituan Secretly Testing a Trillion-Parameter AI Large Model! Currently Only Available to Invited Users
Meituan is secretly testing a new generation of AI large model with a parameter count of trillions, which is entirely trained on a domestic computing cluster, demonstrating confidence in domestic technology. The new model is currently only available to invited users and will promote the application of AI in various fields in the future.
【AiBase Summary:】
📊 Meituan is secretly testing a new generation of AI large model with a parameter count of trillions.
🇨🇳 The model is entirely trained on a domestic computing cluster, showing confidence in domestic technology.
🚀 The new model is currently only available to invited users and will promote AI applications in various fields in the future.
7、Taobao and Tmall Launch AI Identification Technology to Combat Malicious Refunds and Fake Images
Taobao and Tmall have launched AI fake image identification technology, effectively combating malicious refund behaviors and protecting merchants' rights. The technology can accurately identify AI-generated images and forged evidence and provide warnings and dispute resolution support to merchants through the Qiniu Wangwang interface.
【AiBase Summary:】
🛒 AI fake image recognition model is launched, accurately distinguishing between AI-generated images and real image edits.
🛡️ Through the Qiniu Wangwang chat interface, the platform directly adjudicates disputes based on suspected fake images.
📈 Priority is given to high-rated merchants, and it will be expanded to more merchants in the future.
8、Wanxing Technology Launches Wanxing Brain Map AI, Pioneering a New Era of Mind Mapping
Wanxing Technology continues to invest heavily in the field of AI-generated content (AIGC). Its drawing and creativity software "Wanxing Brain Map" has announced the launch of a new feature — Wanxing Brain Map AI. This innovation marks Wanxing Brain Map as the first professional mind mapping software in China to support AI Agent functionality, offering new possibilities for knowledge management and office efficiency. During this upgrade, Wanxing Brain Map AI not only continues to support the familiar hand-drawn mode but also received positive feedback from users during the beta phase, further enhancing the software's interactivity and智能化 level. Through a conversation-driven AI knowledge workbench, users can more conveniently construct and manage mind maps, improving work efficiency and creativity.
【AiBase Summary:】
Support for AI Agent functions in professional mind mapping software
Improve knowledge management and office efficiency
User feedback is positive, enhancing interactivity and智能化 level
