China AI Native Industry Insights – 20260803 – DeepSeek | Alibaba | ByteDance | more
Today’s digest highlights innovative advancements from DeepSeek, Alibaba, and ByteDance. A central theme this week is the enhancement of AI capabilities across different media formats. DeepSeek’s V4-Flash API, now in public beta, promises enhanced agent capabilities and support via its new Responses API, while Alibaba Wan Video introduces a realtime mode for live text-to-image generation. ByteDance’s launch of Seedance 2.5 focuses on a next-generation video creation model. Collectively, these developments underscore the rapid evolution and specialization of artificial intelligence in digital media creation. Discover more in Today’s China AI Native Industry Insights.
1. DeepSeek Launches V4-Flash API in Public Beta with Upgraded Agent Capabilities and Responses API Support
DeepSeek has launched its V4-Flash API in public beta under the build name DeepSeek-V4-Flash-0731. The model retains the same architecture and size as the preview version but has been retrained, delivering agent benchmark scores that significantly surpass V4-Pro-Preview. The release adds native support for the Responses API format and is fully adapted for Codex, with Codex integration currently exclusive to V4-Flash. The update applies only to the V4-Flash API; the V4-Pro API and the models used on DeepSeek’s app and website remain unchanged.
Read more: https://technode.com/2026/07/31/deepseek-puts-v4-flash-api-into-public-beta/
Video Credit: NotebookLM
2. Alibaba Wan Video Launches Realtime Mode for Live Text-to-Image Generation
Alibaba’s Wan video platform has launched a new Realtime mode that allows users to type a text description and watch the image reshape itself in real time with every sentence entered. The feature eliminates the need for a generate button or loading screen, offering a continuous, interactive canvas experience. The update targets creators and users who want a more fluid, conversational approach to AI image generation.
Read more: https://x.com/i/web/status/2082748687533715938
Video Credit: @Alibaba_Wan on X
3. ByteDance Launches Seedance 2.5: A New-Generation Video Creation Model
ByteDance has officially launched Seedance 2.5, its new-generation video creation model. Building on its predecessor, the model brings major advancements across three fronts: long-form storytelling, multimodal referencing, and fine-grained editing. Users can now generate high-quality, 30-second audio-video clips in a single pass, bringing a complete story to life in one take — a significant jump from the roughly 5-to-15-second ceiling of prior models. The model also accepts up to 30 images, 10 video clips, and 10 audio clips as reference material in a single input, substantially improving consistency of characters, style, and motion. Additionally, Seedance 2.5 introduces timestamp-level control, allowing users to make targeted edits to specific audio or video segments without regenerating the entire clip — a notable improvement in both efficiency and controllability. The model is rolling out first on ByteDance’s Jimeng and Doubao Pro platforms, with API access on Volcano Engine’s Ark platform expected to follow.
Read more: https://seed.bytedance.com/zh/seedance2_5
Video Credit: @ByteDanceSeed_ on X
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.