AI Daily: ByteDance Launches Seedance 2.0 Mini; Kimi 2.7 Code High-Speed Version Large Model Officially Launched; DeepSeek Completes Over $7 Billion First Round of Funding

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. Cost per second cut in half, ByteDance launches Seedance 2.0 Mini video generation model

ByteDance's火山引擎 (Volc Engine) launched the Seedance 2.0 Mini video generation model, which is known for its high cost-effectiveness, improving production efficiency and reducing enterprise costs, suitable for scenarios such as e-commerce and marketing content production.

【AiBase Summary:】

🚀 Seedance 2.0 Mini improves video generation speed, twice as fast as the standard fast version.

💰 The cost of generating a second of video is reduced to 0.5 yuan, significantly cutting enterprise costs.

💡 Suitable for high-frequency, large-volume video generation needs, empowering e-commerce and marketing content production.

2. Output speed increased six times, Moonshot officially launched Kimi 2.7 Code high-speed version of the large model

Moonshot released the Kimi 2.7 Code high-speed version, which has an output speed 5 to 6 times faster than the regular version, significantly improving code writing efficiency. In specific scenarios, it can reach up to 260 Token/s, and the price is only twice that of the regular version.

【AiBase Summary:】

🚀 Output speed increased to 5 to 6 times faster than the regular version, significantly improving development efficiency.

💰 The high-speed version API costs only twice that of the regular version, offering high cost-effectiveness.

⚙️ To use the high-speed version, you need to enable the thinking mode; otherwise, it will revert to the K2.6 version.

3. Tencent invests 20 million USD in Lin Junyang's new AI lab

Lin Junyang's AI laboratory received a huge investment from Tencent, Gao Rong Capital, and Sequoia China, showing its huge potential and market attention in the AI field.

【AiBase Summary:】

🧠 Lin Junyang was once the youngest P10-level technical expert at Alibaba, known as one of the "Four Great Models."

💰 Tencent invested 20 million USD, while Gao Rong Capital and Sequoia China each invested 100 million USD, making the lab valued at 2 billion USD.

🚀 The lab has already started preparing for the next round of financing, showing its market popularity and expansion ambitions.

4. Alibaba Cloud QoderWork Launches "Consciousness" Function, Achieving AI Self-Memory and Skill Evolution

Alibaba Cloud QoderWork launched the "Consciousness" function, marking the evolution of desktop-level AI assistants into intelligent agents, with self-memory and skill evolution capabilities, improving long-term collaborative work efficiency, and providing a low-cost, high-privacy personal AI workflow technology demonstration for the industry.

【AiBase Summary:】

🧠 The "Consciousness" function of QoderWork integrates three core modules: memory, reflection, and skill evolution, enabling the AI assistant to have self-growth capabilities.

🔒 Through a hierarchical memory system and selective forgetting mechanism, information security is ensured, and long-term collaborative efficiency is improved.

🚀 The system uses a multi-factor weighted scoring model to achieve efficient skill evolution, while keeping additional costs within 5%.

5. DeepSeek Completes Over $7 Billion First Round Financing: Valuation Exceeds $50 Billion

DeepSeek completed a $7 billion financing through an innovative funding method, with a valuation exceeding $50 billion, while ensuring the founder's control over the company, and attracting significant investments from Tencent and宁德时代 (CATL).

【AiBase Summary:】

💰 DeepSeek completed over $7 billion in financing through an unconventional transaction structure, with a post-money valuation exceeding $50 billion.

🔒 Investor funds entered a limited partnership managed by the CEO, subject to a five-year lock-up period and without voting rights.

🤝 Tencent and CATL plan to invest 1 billion and 500 million RMB respectively, becoming the largest external investors.

6. New "Group Order" Model with AI: OpenRouter Launches Fusion API, Focused on Performance and Cost-Effectiveness Optimization

OpenRouter's Fusion API enhances answer accuracy and quality through a multi-model collaboration mechanism, and performs well in cost control, offering developers a new technological path.

【AiBase Summary:】

🧠 Fusion API enhances answer accuracy and quality through multi-model collaboration.

💰 Fusion API performs well in cost control, offering high cost-effectiveness.

🚀 OpenRouter's Fusion API provides developers with a new technological path.

7. Alipay Launches AI Version of "Abao" for Testing, Driving Mobile Payment Toward "Conversational Service" Transformation

Alipay officially launched the AI version and opened user testing, introducing the "Abao" AI assistant, achieving an upgrade from traditional apps to conversational experiences, while emphasizing fund security and ecological openness, promoting the popularization of AI capabilities.

【AiBase Summary:】

🤖 Alipay launched the AI assistant "Abao," supporting text or voice input, automatically matching mini programs or service interfaces.

🔒 Emphasizing the principle that "AI only handles tasks, fund management remains with users," ensuring fund security.

🤝 Alipay fully opens service access points and protocols, inviting multiple parties to join, promoting the popularization of AI capabilities to all users.

8. Alibaba Releases Qwen-Robot Series Embodied Large Models: Three Models Collaborate to Solve Heterogeneous Robot Adaptation Pain Points

Alibaba released the Qwen-Robot series of embodied intelligence large models, covering three core matrices: operation, mobility, and world models, aiming to solve the adaptation pain points of heterogeneous robots and promote the general deployment of robots.

【AiBase Summary:】

🧠 Qwen-RobotManip achieves rapid adaptation across different hardware through an 80-dimensional unified action representation.

🧭 Qwen-RobotNav unifies language instruction navigation and target search tasks into a single framework.

🧠 Qwen-RobotWorld gives the system the ability to reason about the physical world, predicting and simulating the next action and state.