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://app.aibase.com/zh
1. Xiaomi releases MiMo Claw official version: supports 1,000 continuous tool calls, free time increased to 4 hours
Xiaomi released the official version of MiMo Claw, supporting 1,000 continuous tool calls, with free time increased to 4 hours.
【AiBase Highlights:】
🧠 MiMo-V2.5-Pro supports the MCP tool call protocol, suitable for complex, multi-step task scenarios.
📊 MiMo Claw achieved a task success rate of 63.8% on the ClawEval benchmark, with token consumption reduced by 40% to 60%.
💰 First subscription price is 14.9 yuan/month, discounted subscription price is 19.9 yuan/month or 233.8 yuan/year.
2. Report: Douyin's Dabao daily active users exceed 200 million but revenue is less than a million, ByteDance's AI resources may shift toward enterprise services
The AI product Dabao under ByteDance has daily active users exceeding 200 million, but its revenue is far lower than computing costs, mainly relying on e-commerce commissions. To cope with cost pressures, ByteDance has begun to shift AI resources toward enterprise services and cover some costs through enterprise-level businesses like Seedance.
【AiBase Highlights:】
📱 Dabao has over 200 million daily active users, but daily revenue is less than one million.
💼 ByteDance adjusts AI strategic resource allocation, shifting toward enterprise service lines.
💰 Seedance has an annualized revenue of 2 billion USD, with monthly revenue exceeding 1 billion yuan.
3. Zhipu AI officially opens source GLM-5.2 model: focuses on 1M lossless context and long-range code tasks
Zhipu AI launched the new generation large model GLM-5.2, focusing on code generation and long-range task execution. The model ranks second globally and first among open-source models on Code Arena, and achieves 1M lossless context, improving long-text processing capabilities. In addition, GLM-5.2 introduces thought level control, optimizes the underlying architecture, and supports full-chain development across multiple platforms.
【AiBase Highlights:】
🚀 GLM-5.2 ranks second globally and first among open-source models on Code Arena.
🧠 Achieves 1M lossless context, solving the problem of long-text degradation.
🔧 Introduces thought level control, enhancing programming evaluation performance.
4. WeChat Pay officially launches "AI Dedicated Card": supports Agent closed-loop consumption, main account fully isolated
WeChat Pay launched the "AI Dedicated Card," providing closed-loop consumption capabilities for Agents, ensuring fund security through main account isolation design, and is expected to accelerate the commercialization of the AI ecosystem toward service consumption.
【AiBase Highlights:】
🛒 AI Dedicated Card supports Agent closed-loop consumption, achieving end-to-end service from recommendation to payment.
🔒 Main account fully isolated, ensuring fund security and preventing unauthorized consumption risks.
🚀 It will be opened to more Agent platforms in the future, promoting the commercialization of the AI ecosystem.
5. Cut costs by nearly 40%! Microsoft announces Copilot Cowork AI, directly competing with Claude
Microsoft launched the Copilot Cowork AI product, challenging market competitors with high cost-effectiveness and enterprise-level features, while also focusing on security and flexible billing models.
【AiBase Highlights:】
✅ Microsoft Copilot Cowork is now available globally, offering high cost-effectiveness, with costs 30%-40% lower than Claude Cowork.
🚀 Designed specifically for enterprises, it supports end-to-end execution of complex tasks, allowing users to define goals only.
🔒 Secure and compliant, it inherits the trust boundary of Microsoft 365, and offers flexible billing models.
6. Anthropic urgently halts Claude Agent SDK billing changes, heavy users temporarily continue existing subscription limits
Anthropic has urgently halted the billing changes for the Claude Agent SDK to better support the development needs of subscribed users, reflecting its complex competition between commercialization, computing power management, and developer ecosystems.
【AiBase Highlights:】
🤖 Anthropic pauses the billing changes for the Claude Agent SDK, maintaining existing subscription limits.
💡 Heavy users and developers can continue to enjoy a relaxed weekly standard limit.
🔄 Big model manufacturers face complex competition between commercialization pressure and developer ecosystems.
7. Kunlun Wanjie releases Tian Gong 3.1: introduces new canvas design and multi-Agent parallel workflow
Kunlun Wanjie released Tian Gong 3.1, launching two core capabilities: Skywork Design and Dynamic Workflows, marking the acceleration of large model applications toward complex, multi-agent collaborative graphical workflows.
【AiBase Highlights:】
🎨 Skywork Design enables efficient design for scenarios such as website full site and App prototypes through canvas design.
🤖 Dynamic Workflows support multi-Agent parallel work, improving task processing efficiency.
📈 Tian Gong 3.1 drives AI applications from text interaction to graphical workflows.
8. OpenAI is reported to be preparing to launch a new bidirectional voice model "GPT-Bidi-1"
OpenAI is preparing to launch the new bidirectional voice model GPT-Bidi-1, which achieves a breakthrough upgrade in voice interaction through a bidirectional architecture. It can capture user interruptions in real-time and dynamically adjust semantic output without stuttering, significantly improving the naturalness of real-time voice conversations. Additionally, the model introduces voice intelligence and speed levels, allowing users to balance interaction depth and response speed based on specific tasks.
【AiBase Highlights:】
🧠 GPT-Bidi-1 uses a bidirectional architecture to achieve a breakthrough upgrade in AI voice interaction.
🔄 Supports real-time capture of user interruptions, improving conversation naturalness.
📊 Introduces voice intelligence and speed levels, meeting different task requirements.
