China AI Native Industry Insights – 20260713 – Alibaba | Tencent | ByteDance | more
Explore Alibaba’s new video model, Tencent’s AI storage, ByteDance’s image innovation. Discover more in Today’s China AI Native Industry Insights.
1. Alibaba Releases Wan-Streamer v0.2, an Omni-Modal Model for Real-Time Duplex Video Conversation
Alibaba’s Wan team has released Wan-Streamer, an end-to-end omni-modal understanding and generation model built for real-time full-duplex interaction. The system uses a single native-streaming Transformer architecture capable of simultaneously listening, watching, understanding, and responding with synchronized speech and video at approximately 550ms total latency. Unlike avatar-based systems, Wan-Streamer is not tied to a fixed avatar rig, allowing it to embody any character describable in natural language, including human figures, pets, and anime characters. The v0.2 update raises output resolution to 640×368 at 25 FPS and reduces model-side latency to approximately 200ms.
Read more: https://arxiv.org/abs/2607.04443
Video Credit: NotebookLM
2. Tencent Cloud launches Agent Bucket, a native storage solution for AI agents
Tencent Cloud has officially released Agent Bucket, a native storage solution designed to provide independent cloud space for billions of AI agents. The product introduces a Space concept within buckets to enable native isolation, allowing each agent or user to have independent access credentials, capacity quotas, and read-write rate limits with standard S3 interface compatibility. Agent Bucket integrates features such as file deduplication, POSIX file system semantics, multimodal search, and GooseFS data acceleration for high-concurrency scenarios. The solution is already deployed in Tencent’s WorkBuddy, QClaw, and ClawPro applications, as well as in customer scenarios including ASUS AI PCs and Dongfeng Nissan connected vehicles.
Read more: https://mp.weixin.qq.com/s/Bpz94DgHeYgRi6gU7e8bAA
Video Credit: NotebookLM
3. ByteDance releases Seedream 5.0 Pro multimodal image generation model with advanced editing and multilingual capabilities
ByteDance announced the release of Seedream 5.0 Pro, a multimodal image generation model designed for professional creative workflows. The model features four core capabilities: complex information visualization for data-dense content like infographics and charts, interactive precision editing including point selection, sketch rendering, color and material replacement, realistic photographic and portrait rendering with accurate lighting and texture details, and native support for over ten languages including French, German, Russian, Japanese, Korean, Spanish, and Arabic with proper typesetting rules. The model is now available on Volcano Engine experience center, Doubao app, and Jimeng platform for users to try.
Read more: https://mp.weixin.qq.com/s/COjQ-auPvV9Cc3AFZZxzbg
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.