China AI Native Industry Insights – 20250428 – Alibaba | StepFun Technology | Zhipu AI | more
Explore the launch of Qwen Chat App on iOS and Android for free AI assistance, the release of StepFun AI’s Step1X-Edit for state-of-the-art image editing, and the strategic partnership between Zhipu AI and Shengshu Tech for multimodal innovation. Discover more in Today’s China AI Native Industry Insights.
1. Qwen Chat App Launches on iOS and Android: Free AI Assistance for Creativity and Collaboration
🔑 Key Details:
– Qwen Chat App is now available for both iOS and Android users, offering a free, user-friendly experience designed to enhance creativity and collaboration.
– Supports natural conversation-based interaction, helping users brainstorm ideas, solve problems, and explore possibilities with ease.
– Simple onboarding — scan a QR code to download and start using instantly, with no complicated setup required.
– Currently available in selected markets, with broader rollout planned.
– Built to assist users across creative, professional, and everyday tasks, powered by cutting-edge AI technology.
💡 How It Helps:
– Creatives: Instantly brainstorm, refine ideas, and explore new possibilities through natural conversations.
– Students and Professionals: Get quick assistance for writing, problem-solving, and research without switching platforms.
– Teams: Foster collaboration and innovation by integrating an AI partner into everyday workflows.
– New Users: Easy access to advanced AI assistance without the need for technical expertise.
🌟 Why It Matters: Qwen Chat democratizes access to advanced conversational AI, making it easy for individuals and teams to enhance their creativity, productivity, and problem-solving abilities. By offering a free, intuitive app experience, it opens the door for more people to explore the potential of AI-powered collaboration in their daily lives.
Original article: https://x.com/Alibaba_Qwen/status/1915761990703697925
Video Credit: Qwen (@Alibaba_Qwen on X)
2. StepFun AI Releases Step1X-Edit: Open-Source Image Editing Model with SOTA Performance
🔑 Key Details:
– 19B Parameter Model: Step1X-Edit combines a 7B MLLM and 12B DiT architecture, achieving state-of-the-art performance among open-source image editing models.
– Comprehensive Editing: Supports 11 types of image editing tasks including text replacement, style transfer, material transformation, and portrait retouching.
– Technical Innovation: Implements a decoupled MLLM+Diffusion architecture that separates language understanding from high-fidelity image generation.
– Benchmark Leader: Outperforms existing open-source models on GEdit-Bench metrics, approaching GPT-4o and Gemini 2.0 Flash performance levels.
💡 How It Helps:
– Content Creators: Natural language instruction support eliminates the need for templates, enabling intuitive editing through conversational prompts.
– E-commerce Platforms: Identity consistency preservation maintains facial features and posture, ideal for virtual models and product images.
– Graphic Designers: Precise region-level control for targeted editing of text, materials, and colors while maintaining overall image style.
– AI Researchers: Open-source implementation available on GitHub, HuggingFace, and ModelScope with technical documentation.
🌟 Why It Matters:
Step1X-Edit represents a significant advancement in democratizing professional-grade image editing through AI. Its ability to understand complex instructions, preserve identity, and maintain image consistency addresses key limitations of previous models. By achieving commercial-grade performance in an open-source package, StepFun AI is enabling wider innovation across industries while challenging proprietary solutions from major AI labs.
Original Chinese article: https://mp.weixin.qq.com/s/2gGraRc5zpCa5bTms7kVxw
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2F2gGraRc5zpCa5bTms7kVxw
Video Credit: The original article
3. Zhipu AI and Shengshu Tech Form Strategic Partnership on Multimodal AI Innovation
🔑 Key Details:
– Zhipu AI and Shengshu Technology announced a strategic partnership on April 27, 2025, combining expertise in large language models and multimodal generation.
– Zhipu’s MaaS platform will integrate Shengshu’s Vidu API, enhancing video generation capabilities for developers and enterprise clients.
– The collaboration targets sectors like government services, internet, cultural tourism, advertising, animation, and media with customized AI solutions.
– Both companies originate from Tsinghua University, leveraging Zhipu’s GLM series and Shengshu’s multimodal strengths.
💡 How It Helps:
– AI Developers: Gain seamless access to integrated video generation via a unified platform.
– Enterprise Providers: Deliver industry-specific AI solutions with reduced complexity.
– Media Creators: Enhance content production workflows with advanced video generation tools.
– Public Sector: Tailored AI applications to meet government and civic needs.
🌟 Why It Matters:
The partnership consolidates two major players in China’s AI space, merging language and multimodal capabilities into a more powerful offering. It signals a shift toward industry-focused solutions, potentially boosting China’s competitiveness in practical AI deployment across key sectors.
Original Chinese article: https://mp.weixin.qq.com/s/r0ra519SdASJKqYqgltVrw
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2Fr0ra519SdASJKqYqgltVrw
Video Credit: The original article
4. Moonshot AI Releases Kimi-Audio-7B: New Open-Source Audio Foundation Model
🔑 Key Details:
– Universal Audio Capabilities: State-of-the-art model for speech recognition, audio Q&A, captioning, emotion recognition, and speech conversation in a single framework.
– Open-Source Release: Pretrained model weights, instruction-tuned model, comprehensive evaluation toolkit, and technical report now available.
– Advanced Architecture: Features hybrid audio input system, LLM core with parallel heads, and efficient chunk-wise streaming detokenizer.
– Performance Leader: Achieves superior results across multiple benchmarks including LibriSpeech, MMAU, and VoiceBench.
💡 How It Helps:
– AI Researchers: Comprehensive evaluation toolkit enables reproduction of results and standardized benchmarking across models.
– Speech Application Developers: Ready-to-implement model with documented inference examples for quick integration.
– Audio Engineers: Hybrid approach combining discrete semantic and continuous acoustic tokens enhances audio quality and understanding.
🌟 Why It Matters:
Kimi-Audio represents a significant advance in unified audio processing, breaking down traditional task-specific boundaries to create a true foundation model. The open-sourcing of both model and evaluation tools addresses the critical challenge of inconsistent benchmarking in audio AI, potentially accelerating development across the field.
Original article: https://github.com/MoonshotAI/Kimi-Audio
Video Credit: The original article
5. Tsinghua University Establishes AI Hospital to Transform Healthcare
🔑 Key Details:
– New Institution: Tsinghua University officially launched its AI Hospital (Tsinghua AI Agent Hospital) on April 26, 2024, with President Li Luming and Vice President Wang Hongwei attending the ceremony.
– Phased Implementation: The AI Hospital will initially operate as a system within Beijing Tsinghua Changgeng Hospital, focusing on general medicine, ophthalmology, radiology, and respiratory medicine.
– Advanced AI: The “Zijing AI Doctor” testing system went online in November 2023, establishing technological foundations for the AI Hospital through closed-loop medical virtual environments.
💡 How It Helps:
– Medical Practitioners: AI-assisted decision-making tools increase diagnostic precision and service efficiency while reducing operational costs.
– Medical Educators: The hospital serves as an educational platform for training a new generation of “AI-collaborative doctors” with both medical expertise and AI literacy.
– Healthcare Administrators: Addresses the shortage of primary care physicians and enables more efficient distribution of quality medical resources.
🌟 Why It Matters:
This initiative represents a fundamental shift from the traditional “hospital+AI” model to an AI-integrated medical institution designed from the ground up. By leveraging Tsinghua’s interdisciplinary advantages in engineering and medicine, the AI Hospital aims to revolutionize healthcare delivery, making high-quality medical services more affordable and accessible. This aligns with China’s “Healthy China 2030” strategy while positioning Tsinghua at the forefront of AI-powered healthcare transformation.
Original Chinese article: https://mp.weixin.qq.com/s/n-87KYCszONioyJrmqOJ5A
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2Fn-87KYCszONioyJrmqOJ5A
Video Credit: The original article
That’s all for today’s China AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.