China AI Native Industry Insights – 20261106 – Tencent | Moonshot AI | Baidu | more
Today’s digest highlights significant advancements by Tencent, Moonshot AI, Baidu, and Ant Group in the AI landscape. A common thread among these announcements is the emphasis on enhancing model capabilities, whether through improved vision features, open-weight access, or exceptional benchmark performance. Tencent’s Hunyuan Model Hy4 brings new vision capabilities to the table, while Ant Group’s Ling-3.2-flash impressively features 560 billion parameters, surpassing Opus 5 on crucial benchmarks. Moonshot AI introduces the Kimi K3.1 Pro with open-weight, and Baidu unveils its ERNIE 5.5 Flash Base, showcasing the evolving focus towards sparse MoE language models. Discover more in Today’s China AI Native Industry Insights.
1. Tencent Releases New Hunyuan Model Hy4 with Vision Capabilities
Tencent’s Hunyuan team launched a new model called Hy4 that adds vision capabilities to its lineup. The company says the model ranks among the leading open models available. Hy4 is positioned for use across finance, frontend development, gaming, office work, and research tasks. Tencent states the model was built with cybersecurity and biosecurity considerations in mind.
Read more: https://hy.tencent.ai/research/hy4?langVersion=en
Video Credit: The original article
2. Moonshot AI releases open-weight Kimi K3.1 Pro model
Moonshot AI has released the model weights for Kimi K3.1 Pro, making the model openly available. The company stated the release reflects its continued commitment to advancing open intelligence. The announcement was shared via the official Kimi Moonshot account. No additional technical specifications were disclosed in the announcement itself.
Read more: https://huggingface.co/moonshotai/Kimi-K3.1-Pro
Video Credit: NotebookLM
3. Baidu Releases ERNIE 5.5 Flash Base, a Sparse MoE Language Model
Baidu released ERNIE 5.5 Flash Base, an efficient sparse mixture-of-experts base language model. The model has 280 billion total parameters with 14 billion activated per token, and was pre-trained on 33.5 trillion tokens across multiple domains and languages. It achieves state-of-the-art results on multiple benchmarks among similarly sized base models and supports up to 1 million tokens of context for fine-tuning and inference. The weights are released directly from pre-training, with no mid- or post-training applied, and are available in FP8 and BF16 formats.
Read more: https://ernie.baidu.com/release/ernie-5-5-flash-base/
Video Credit: NotebookLM
4. Ant Group Unveils Ling-3.2-flash, a 560B-Parameter Multimodal Model Outperforming Opus 5 on Key Benchmarks
Ant Group’s Ling AGI team introduced Ling-3.2-flash, a native multimodal model with 560 billion total parameters and 25 billion active parameters. The model emphasizes stronger complex visual understanding, long-horizon multimodal agent capabilities, 3D understanding and interaction, and medical intelligence. According to release evaluations shared by the team, Ling-3.2-flash outscored Opus 5 (max) on OmniDocBench1.5 (92.21 vs 90.56), ZEROBench Sub (43.71 vs 39.52), and Vision2Web (85.86 vs 80.20). The announcement was made via the AntLingAGI account on November 5, 2026.
Read more: https://x.com/AntLingAGI/status/2118428246664761379
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.