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 technical trends and innovative AI product applications.
Hot AI products Click to learn more:https://news.xixuncloud.com/zh
1. Under $400, you can train it yourself! Hugging Face launches the Microduck, a miniature open-source bipedal robot
The miniature open-source bipedal robot Microduck launched by Hugging Face is not only cute in appearance but also integrates cutting-edge technologies such as reinforcement learning, Sim2Real, and real-time control, becoming an important platform for research on embodied intelligence.
【MLA-AI Summary:】
🤖 Microduck is an open-source miniature bipedal robot with various practical skills.
🧠 It uses reinforcement learning and Sim2Real technology to achieve transfer from simulation to reality.
🔧 The software architecture is written in Rust and divided into multiple independent services to ensure reliability.
More details: https://pollen-robotics.com/microduck/
2. Sold out immediately! Hy4preview Agent's capabilities have been greatly upgraded, and Tencent has urgently expanded its cluster
Tencent's Hy4preview caused high concurrency demand upon launch due to significant improvements in agent capabilities, leading to task queue backlogs. Tencent has urgently expanded its inference cluster and introduced a transitional alternative solution to meet user needs.
【MLA-AI Summary:】
🚀 The upgrade of Hy4preview Agent capabilities has sparked a热潮 among users and developers.
🔄 Tencent has urgently expanded its inference cluster to handle high concurrency demands.
📅 The transitional alternative solution will be extended until September 30, 2026.
3. iFLYTEK Will Launch the Starfire X2.5 General Large Model
iFLYTEK is about to launch the Starfire X2.5 general large model, which includes edge models and base models, enhancing capabilities such as intelligent agent interaction, mathematical calculation, and code generation. It plans to release new models based on fully domestic computing power to promote technological innovation.
【MLA-AI Summary:】
🧠 The Starfire X2.5-4B and Starfire X2.5-1.7B edge models support a context window of up to 1M tokens, enhancing intelligent agent interaction and mathematical calculation capabilities.
💻 The Starfire X2.5 base model has a scale of 293B, with deep upgrades in core technologies such as code generation and intelligent agent collaboration.
🚀 iFLYTEK plans to launch a new main general large model based on fully domestic computing power, continuously promoting the self-innovation of underlying technologies.
4. China's First Full AI-Generated Long Drama "The Later Journey to the West" Premieres Today, Introducing the "Review and Broadcast as You Go" Model
"The Later Journey to the West," as China's first full AI-generated long drama, not only continues the plot clues of the classic "Journey to the West" in content but also achieves industry breakthroughs in production mode by adopting the innovative "review and broadcast as you go" approach, providing a new exploration for AI-era film and television content production.
【MLA-AI Summary:】
🍎 China's first full AI-generated long drama "The Later Journey to the West" officially premiered, marking a new chapter in AI film and television production.
📺 The show adopts the "review and broadcast as you go" model, significantly shortening the production and broadcasting cycle and improving efficiency.
📖 The series is adapted from the Ming and Qing dynasty novel "The Later Journey to the West," building a complete "Later Journey to the West" narrative system.
5. OpenAI Completes Pre-training of New 10 Trillion Parameter Model Bel
OpenAI successfully completed the pre-training of the large-scale pre-training model named "Bel," with a parameter count reaching 10 trillion, surpassing Astra and establishing a solid barrier in computing power and self-developed chip technology.
【MLA-AI Summary:】
🧠 The Bel model has a parameter count of 10 trillion, achieving a technological breakthrough.
🚀 The Bel model surpasses Astra in performance and architecture and has self-recovery capabilities.
💡 OpenAI has built a technical barrier through self-developed chips and computing power reserves.
6. ByteDance Increases Incentives for Core AI Teams, Doubao Stock Price Rises 14.6% to Boost Big Model Strategy
ByteDance recently adjusted its equity incentive policy for Doubao stock, increasing the price and distribution of Doubao stock to bind the core AI big model team more deeply by enhancing the value and flexibility of equity in key business segments.
【MLA-AI Summary:】
📌 The price of Doubao stock was raised from $14.85 per share to $17.02 per share, an increase of about 14.6%
📌 The value of some employees' Doubao stock can reach around 5% of their total compensation package
📌 Different levels of employees have different full-configuration exchange ratios, including 20%, 30%, and other tiers
7. OpenClaw 2.0 Officially Released: Reconstructing the Bottom Architecture, Supporting Multi-user Shared Cloud Sessions
The release of OpenClaw 2.0 marks the evolution of AI agents from single-machine auxiliary tools to collaborative and autonomous productivity infrastructure, improving the deployment efficiency of intelligent agents.
【MLA-AI Summary:】
🔧 Reconstructing the bottom architecture to improve system performance and stability.
👥 Support multi-user shared cloud sessions to enable team collaboration and information sharing.
🚀 Simplify the installation process and provide an easy-to-use experience right out of the box.
8. One Idea, One AI Team: Qwen Creation Agent Teams Launched, Screenplay, Directors, and Artwork All Covered
Qwen Creation launched the Agent Teams feature, enabling full-process automated production from idea to final product through multi-agent collaboration, significantly lowering the barrier to professional content production.
【MLA-AI Summary:】
🎬 Qwen Creation launched the Agent Teams feature, realizing a multi-agent collaboration model.
🧠 First access to the Wan3.0 video generation model to enhance generation quality and speed.
🔄 Covers AI short dramas, advertisements, games, and other content scenarios, reducing the barrier to content production.