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 to help you understand technology trends and innovative AI product applications.
Fresh AI products Click to learn more:https://news.xixuncloud.com/zh
1、World Labs launches the world's first multimodal world model Atlas, achieving cinematic "bullet time" effects with pixel-level camera control
The Atlas model launched by World Labs is the world's first multimodal world model. Its core breakthrough lies in generating high-quality images and videos through pixel-level camera control and achieving precise 3D space reconstruction. The model not only can generate 3D models from multiple images but also supports image generation and panoramic views from text, demonstrating strong world generation, reconstruction, and simulation capabilities.
【MLA-AI Highlights:】
🖼️ Atlas model can generate images and video frames with pixel-level precise camera control and perfectly reconstruct in 3D space.
🎥 The model supports outputting clear 3D models from one or multiple input images, with reconstruction results surpassing many top open-source models.
📽️ Atlas has spatiotemporal simulation capabilities, allowing video re-composition to enhance dramatic visual effects, and supports image generation from text and 360-degree panoramas.
2、Tencent WorkBuddy announces the launch of its open platform, fully opening up Agent base capabilities
Tencent WorkBuddy announced the launch of its open platform, fully opening up Agent base capabilities, aiming to connect more capabilities and enter more scenarios, becoming a platform that carries all tools.
【MLA-AI Highlights:】
🤖 Tencent WorkBuddy's open platform officially launched, with over a hundred ecosystem partners introduced in the first batch.
💼 Nine co-branded smart hardware products debuted, covering multiple industry fields.
🚀 A vice president of Tencent Cloud stated that WorkBuddy will create an operating system for the Agent era.
3、Cyberspace Administration of China intensifies efforts to crack down on AI face swapping and modified classics; Doubao, Yuanbao, Qianwen, and Wenxin Yinyan strictly limit the output of illegal content
The Central Cyberspace Administration launched the "Clear Brightness - Crackdown on AI Application Chaos" special operation, targeting the abuse of AI technology for in-depth governance, including clearing illegal and irregular information, investigating illegal accounts, and handling illegal websites and applications. Local cyberspace departments have introduced precise governance measures while mainstream platforms have enhanced technological empowerment to improve intelligent detection efficiency. In addition, leading large model platforms have strictly reviewed original content, and app stores have tightened developer access standards to prevent the spread of AI-generated illegal content.
【MLA-AI Highlights:】
🧠 The abuse of AI technology is being heavily rectified, including the creation of false information and the spread of vulgar content.
🛡️ Local cyberspace departments have introduced precise governance measures to enhance the ability to identify and handle illegal content.
🔒 Through source review and platform technological upgrades, the spread of AI-generated illegal content is effectively curbed.
4、Doubao Mobile Phone to be released in September: Named Nubia NaviX Ultra, called the world's first AI agent flagship phone
The Nubia NaviX Ultra, as the world's first AI agent flagship phone, achieves a fundamental difference from traditional AI phones by embedding a large model agent into the system's core. Its powerful intelligent operation capabilities and privacy security guarantees offer users a brand new experience.
【MLA-AI Highlights:】
📱 AI-native phone architecture: Large model agents embedded in the system’s core for a more efficient user experience.
✈️ Intelligent operation capabilities: AI can automatically complete complex tasks like booking tickets without user intervention.
🔒 Privacy security: All operations and memories are completed locally, without uploading data to the cloud.
5、Meta introduces its first real-time audio perception model: Can separate and transcribe speech from over 20 people simultaneously, at $3 per thousand minutes
Meta's Muse Voice Transcribe is the first real-time audio perception model, supporting streaming voice recognition, speaker separation, and endpoint detection, suitable for meeting notes and call subtitles. The model can process recordings from more than 20 speakers and supports over 70 languages, while introducing an adaptive delay mechanism to balance speed and accuracy.
【MLA-AI Highlights:】
🎙️ Real-time audio perception model, supports streaming voice recognition while speaking.
👥 Can separate over 20 speakers and automatically transcribe each track.
🌐 Supports over 70 languages, with an adaptive delay mechanism to improve accuracy.
6、The world's most powerful model enters a price war! Anthropic launches Fable 5.1: Performance soars and costs drop by 40%
Anthropic launched its new flagship large model, Fable 5.1, which has significant improvements in performance and cost. Fable 5.1 not only outperforms its predecessor and competitors in various benchmark tests but also significantly reduces usage costs, especially in complex agent-based tasks, where cost savings can reach up to 45%. This marks that the "arms race" in AI large models is extending comprehensively toward extreme cost-effectiveness.
【MLA-AI Highlights:】
🔥 Fable 5.1 outperforms its predecessors and competitors in multiple benchmark tests.
💰 Usage costs decreased by 25%, with cost savings of up to 45% for complex tasks.
🚀 Provides strong technical support for Anthropic's IPO, possibly breaking the fundraising record.
7、Battle between OpenAI and Anthropic! Google Gemini 3.8 Flash internal test code name "Skimaki" to be released as early as Wednesday
Google made a key product update in the field of artificial intelligence, launching the Gemini 3.8 Flash model, which significantly improves programming capabilities, while the DeepMind team underwent personnel adjustments, and Google also increased investment in reinforcement learning.
【MLA-AI Highlights:】
🧠 Google launched the Gemini 3.8 Flash model, internally named "Skimaki," enhancing programming capabilities.
🔄 The DeepMind team underwent personnel changes, with Koray Kavukcuoglu taking over as Senior Vice President, requiring faster execution.
🚀 Google is shifting more R&D time and computing power to the reinforcement learning phase and has recruited Barret Zoph to lead technical challenges.
8、Google AI photo editing tool Pics officially launched: Edit pictures in Docs and Slides with just one click
Google launched the AI image design and editing tool Google Pics, which users can use on the web version or in Google Docs and Slides, supporting material uploads, object segmentation, and text modification functions, with personal and enterprise users gradually gaining access rights.
【MLA-AI Highlights:】
🖼️ Users can upload materials or input prompts to generate images, supporting direct editing in Google Docs and Google Slides.
📐 The tool has precise object segmentation capabilities, allowing users to adjust individual objects or text elements in images separately.
👥 Personal and enterprise users can gradually gain access rights through specific accounts or packages.