AI Daily: Alibaba Launches Open-World Model Happy Oyster; Claude Opus 4.7 to Be Released Soon; Tencent Opensources Huayuan 3D World Model 2.0

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 learn about innovative AI product applications.

Fresh AI products Click to learn more:https://app.aibase.com/zh

1. Alibaba ATH Launches Open-World Model Happy Oyster, Opening Real-Time Interaction Experience Applications

The Alibaba ATH team has launched a new open-world model called Happy Oyster, which has real-time interaction and creation capabilities, further expanding its technical layout in the generative AI field. At the same time, its HappyHorse-1.0 model achieved global first place in the Video Edit direction ranking, demonstrating strong multimodal visual processing capabilities. The release of Happy Oyster reflects the industry's trend towards dynamic and interactive environments and provides higher-dimensional productivity tools for game development, virtual simulation, and creative design.

AiBase Highlights:

🧠 Alibaba ATH team launched a new open-world model Happy Oyster, supporting real-time interaction and creation capabilities.

📊 In the latest Video Edit direction ranking, HappyHorse-1.0 model ranked first globally with 1299 points, showing deep technical strength.

🚀 The release of Happy Oyster marks the evolution trend of generative AI technology from static content generation to dynamic interactive environments.

Details link: https://www.happyoyster.cn/

2. Anthropic to Release Claude Opus 4.7 and New AI Design Tools

Anthropic is set to release Claude Opus 4.7 and new AI design tools, drawing market attention.

AiBase Highlights:

🧠 Claude Opus 4.7 is an incremental upgrade over 4.6, bringing enhanced coding capabilities and extended context windows.

🎨 Anthropic is preparing to launch an AI-driven design tool that directly competes with Gamma and Google Stitch.

💰 Anthropic's valuation has reached $80 billion, with annual revenue increasing from $9 billion to $30 billion.

3. Tencent Opens Source Yuan 3D World Model 2.0, Supporting One-Click Generation of Editable 3D Space Assets

Tencent Yuan team officially released and open-sourced the "Yuan 3D World Model 2.0," which achieved significant upgrades in the multimodal world model, capable of generating interactive and editable 3D assets, supporting multiple formats for export and integration with mainstream game engines. Its technical architecture unifies the process of spatial understanding, generation, and reconstruction, with core upgrades including HY-Pano-2.0 model, self-developed spatial Agent technology, and WorldStereo mechanism, significantly reducing the construction threshold for game level prototypes and embodied intelligent simulation environments.

AiBase Highlights:

🧠 HY-World2.0 has made a leap from generating videos to generating interactive and editable 3D assets.

🚀 Supports multiple format exports and seamless integration with mainstream game engines.

🌐 Through SOTA-level spatial consistency memory mechanisms, it has brought AI-driven 3D content creation to the practical stage.

4. Google Releases Strongest TTS Model, Supporting Nearly 70 Languages

Google's Gemini-TTS model has made significant breakthroughs in the text-to-speech field, allowing users to control the emotion, rhythm, and style of speech through prompts, making the voice more natural and delicate. At the same time, the model supports about 70 languages and can automatically identify the language, providing an efficient solution for multilingual content voice conversion.

AiBase Highlights:

✨ Gemini-TTS supports precise control of speech emotion, rhythm, and style through prompts.

🌍 Supports about 70 languages, automatically identifies language without manual annotation.

🚀 Works with other audio models in the series to enhance real-time dialogue and multimodal interaction experiences.

5. Ant Group's Lingbo Technology Open Sources LingBot-Map: Supports Real-Time Streaming 3D Reconstruction with a Single Camera

Ant Group's Lingbo Technology open-sourced the streaming 3D reconstruction model LingBot-Map, which can estimate camera pose and reconstruct 3D scene structure in real-time using only a single regular RGB camera during video capture, providing efficient, stable, and continuous online mapping capabilities for robot navigation, autonomous driving, and AR hardware.

AiBase Highlights:

📱 Supports real-time estimation of camera pose and 3D scene structure using a single regular RGB camera.

🚀 Uses a streaming processing architecture to achieve real-time interaction while receiving images.

🌐 Open-source model lowers the hardware barrier for high-precision 3D perception, accelerating the development of embodied intelligent devices.

Details link: https://huggingface.co/robbyant/lingbot-map ModelScope: https://www.modelscope.cn/models/Robbyant/lingbot-map

6. Mango TV Membership Exceeds 75.6 Million, Self-Developed Large Model Applied to Over 30 Shows, Hunan Radio and TV's AI Transformation Performance Report Is Here

Hunan Broadcasting System showcased its impressive performance report at the 13th China Network Audio-Visual Conference, including Mango TV's effective membership exceeding 75.6 million and the practical application of its self-developed Mango Large Model in over 30 shows, significantly improving production efficiency.

AiBase Highlights:

🍊 Mango TV's effective membership exceeded 75.6 million, showing its competitiveness in the streaming market.

🤖 Hunan Radio and TV's self-developed Mango Large Model has孵化ed over 80 intelligent agents and applied to more than 30 shows, improving production efficiency by over 30%.

💡 Mango TV's AI transformation is replicable, demonstrating successful attempts by traditional broadcasting institutions in technological application.

7. Over 200,000 AI Servers Exposed to Attack Risks, MCP Exposed with Severe Design Flaws

A security report revealed serious design flaws in Anthropic's MCP protocol, potentially exposing over 200,000 AI servers to remote code execution risks. The vulnerability originated from the STDIO interface of the MCP SDK, allowing any incoming operating system commands to be executed without verification or warnings. Although OX Security had reported this to Anthropic, it had only updated its security documentation and did not modify the architecture, causing concern among developers.

AiBase Highlights:

🧠 MCP architecture design flaws could lead to remote code execution, affecting over 200,000 AI servers.

🛡️ The vulnerability involves Anthropic's MCP protocol, covering 11 programming languages, with widespread risk.

⚠️ Anthropic did not make architectural changes, only updated the security documentation, causing developer concerns.

8. OpenAI Releases Updated Agents SDK to Help Enterprises Build Safer Intelligent Agents

OpenAI has made a major update to its Agents SDK, aiming to help enterprises build intelligent agents more securely and efficiently. This update introduced a sandbox feature, allowing intelligent agents to run in a controlled computing environment, effectively protecting the overall security of the system. Additionally, the new version of the SDK provides a built-in framework that supports the use of files and approved tools within the workspace, enhancing development flexibility. OpenAI stated that the new Agents SDK features will be available to all customers via API and will follow standard pricing. Currently, the new framework and sandbox features will be launched first in Python version, with TypeScript support to follow in subsequent versions, and plans to introduce more agent features in the future.

AiBase Highlights:

🛡️ OpenAI updates the Agents SDK, introducing sandbox features to improve the security of intelligent agents.

🔧 The new SDK provides a built-in framework that supports the use of files and approved tools, enhancing development flexibility.

📈 The API will be available to all customers, with Python and TypeScript versions being rolled out sequentially.