2025.5.22 CHINA AI TOOLS TRENDS NEWS: Skywork Super Agents Leads GAIA Ranking with Multi-Agent Innovation
1. Lovart & Medeo: Dual AI Agents Redefine Design and Video Automation
Two domestic AI agents, Lovart and Medeo, launched simultaneously on May 21-22. Lovart enables end-to-end design workflows (logos, 3D rendering, video effects) through conversational interfaces, integrating Keling and Suno models. Medeo automates short video production from text/articles, generating配音, editing, and covers. Both tools signal China's shift toward task-oriented AI agents.
2. Tencent Upgrades Hunyuan Ecosystem for Industrial AI
At Tencent Cloud's AI Summit (May 21), the Hunyuan model matrix received full upgrades including an industrial AIGC engine. The game visual generation platform now accelerates character/scene creation by 10-30x, while the TurboS model ranks Top 8 globally in LLM benchmarks.
3. Skywork Super Agents Dominates Global AI Research
Kunlun Wanwei launched Skywork Super Agents on May 22, featuring 5 expert agents for office scenarios. It topped Meta's GAIA benchmark (82.42 score), surpassing OpenAI in generating research reports with data charts and authoritative citations. The PPT agent produces presentation decks via deep search integration.
4. Huawei-UBTECH Collaboration Accelerates Humanoid Robotics
Huawei and UBTECH signed a strategic partnership (May 20) to develop AI-powered humanoid robots using大模型 and 5G. Initial applications target manufacturing/logistics, with plans to achieve 10,000-unit annual production capacity by Q3 2025.
5. Midea Tests Humanoid Robots in Manufacturing
Midea Group deployed self-developed humanoid robots at Jingzhou factory (May 22) for equipment inspection and assembly tasks. The robots will enter retail demo phases by mid-2025, showcasing "AI+manufacturing" integration.
6. Unitree Robotics Sees Surging Orders with Full Self-Developed Tech
Unitree announced (May 22) exponential growth in humanoid robot orders, achieving 100% self-developed core components. The Hangzhou-based company plans to expand R&D teams across algorithms, hardware, and testing.
7. Feishu's Enterprise AI Q&A System Boosts Efficiency
ByteDance's Feishu launched "Knowledge Q&A" (May 21), an AI tool combining cloud-based models with enterprise data. It resolves 90% internal queries within 200ms, featuring permission controls and triple data encryption for secure corporate adoption.
8. Tencent CodeBuddy Revolutionizes Developer Workflow
Tencent revealed (May 21) that 85% of programmers now use CodeBuddy, reducing coding time by 40% with AI-generated code exceeding 40% ratio. The tool supports 200+ languages and WeChat Mini Program development, achieving 300% efficiency gains.
9. VolcEngine's Podcast Model Enables Instant Voice Duets
ByteDance's VolcEngine released Doubao·Podcast (May 22), a streaming architecture converting text to conversational audio within seconds. Applied in文旅 guides and AI education, it supports interactive content creation with emotional tone control.
10. DeepSeek's Coding Model Tops Open Source Charts
DeepSeek-Coder v3 achieved SOTA performance on HumanEval benchmarks (May 21), featuring enhanced code completion and debugging. Its 33B parameter model now leads GitHub's trending repositories with 370k+ downloads.
11. Alibaba's Wanxiang-VACE 2.1 Enhances Video Editing
The upgraded open-source model (May 21) supports 720P resolution and multimodal inputs, improving video inpainting accuracy by 18%. It unifies text/image/video processing for streamlined creative workflows.
12. iFlytek Spark Desk Integrates Real-Time Medical AI
iFlytek updated Spark Desk (May 22) with hospital EHR analysis, achieving 94% accuracy in preliminary diagnosis. The tool reduces clinicians' documentation time by 60% through smart voice-to-text conversion.
13. SenseTime Unveils AI City Brain 4.0
Launched on May 22, this urban management system uses 50,000+ IoT devices for real-time traffic/pollution monitoring. It reduces emergency response time by 35% through multi-agent coordination.
14. Baidu PaddleOCR 3.0 Boosts Recognition Accuracy
Baidu's open-source OCR model (May 22) improved text recognition accuracy by 13%, supporting 200+ languages. New features include table reconstruction and handwritten formula recognition for educational applications.
15. Xiaomi's MiMo 2.0 Enhances On-Device AI
The upgraded 7B parameter model (May 22) achieves 400ms latency on smartphones, powering real-time camera scene recognition and IoT coordination. It reduces cloud dependency while maintaining GPT-4 level reasoning capabilities.
See More Content about CHINA AI TOOLS
