China AI Native Industry Insights – 20261110 – Baidu | Qoder | ByteDance | more
Baidu, Qoder, and ByteDance make headlines today with significant advancements in AI technologies. The broader theme highlights the increasing capabilities of AI in creative and research domains, emphasizing automation, anonymity, and enhanced rendering. Baidu’s ERNIE Evolve offers automated AI research, while Qoder’s Qoduck introduces a model that excels in 3D rendering and front-end development with anonymity features. ByteDance’s Seedream 5.5 excels in image generation with the capability of multi-round editing and supports native RGBA transparency. These developments reflect the growing versatility and specialization of AI models in diverse technological areas. Discover more in Today’s China AI Native Industry Insights.
1. Baidu Introduces ERNIE Evolve, an Automated AI Research System
Baidu announced ERNIE Evolve, an automated research system that selects research directions, modifies code, and runs experiments toward a given research goal. Improvements are kept only after passing independent evaluation and review, with each round informing the next. Baidu reported state-of-the-art results on NanoChat AutoResearch and NanoGPT Speedrun benchmarks, along with a 33.7 percent increase in deterministic GPU inference throughput and a 14.9 percent reduction in single-request decoding latency on ERNIE 5.5 Flash. Baidu described the system as a step toward recursive self-improvement and plans to extend the approach to training systems, model capabilities, and evaluation.
Read more: https://ernie.baidu.com/research/posts/ernie-evolve/
Video Credit: NotebookLM
2. Qoder launches Qoduck, a new anonymous AI model with advanced 3D rendering and frontend development capabilities
Qoder released Qoduck, a new anonymous AI model available in its desktop application on November 7, offering free daily credits for personal users. A developer tested Qoduck with 40 generation tasks and found it particularly strong in 3D graphics, creating complex WebGL2 projects like black holes with gravitational lensing and perpetual marble machines without using frameworks like Three.js. The model demonstrates solid frontend development skills, especially when paired with the frontend-rails skill that enforces design system consistency and multi-viewport testing. Qoduck outputs detailed delivery reports documenting its testing process and shows performance comparable to other leading AI coding models.
Read more: https://mp.weixin.qq.com/s?__biz=MzE5ODI2NzI5Nw==&mid=2247494264&idx=1&sn=37e5dd9aa20e4b899ed92b374496bd83&chksm=9703564a4e016ac9026c20bb6f35eea4098e7043c300989674ea2f519e707399e742b9ca5bff&scene=0&xtrack=1#rd
Video Credit: The original article
3. ByteDance releases Seedream 5.5 image generation model with multi-round editing and native RGBA transparency support
ByteDance officially released Seedream 5.5, a new generation image creation model with three core capability upgrades. The model supports pixel-level preservation in non-edited areas during multi-round editing, enabling up to 100 consecutive editing rounds on the same work while reducing detail loss and quality degradation. It natively generates RGBA transparent images with Alpha channel, eliminating the need for post-production background removal. The model also improves realistic rendering of lighting, materials, and skin textures, and significantly enhances text rendering accuracy for complex characters and small dense text. Benchmark results show Seedream 5.5 leads mainstream models in text structure and text-to-image realism, and the model is now rolling out on platforms including Jimeng, Doubao, Doubao Work, and Volcano Engine.
Read more: https://mp.weixin.qq.com/s?__biz=MzkzMDY5MzYxNg==&mid=2247494758&idx=1&sn=25bbe45475edfa51ba3d0db95efda781&chksm=c3485ef2cde949c37656975a21f29a1608a307b8c6acb195adb48be35f9a0a1925a51d281e46&scene=0&xtrack=1#rd
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.