DeepSeek, once nicknamed the large-model “price butcher”, is sheathing the knife. At 00:00 on 17 August, its new API prices took effect. Founder Liang Wenfeng is walking back the rock-bottom pricing that built DeepSeek’s brand. (Source: Sohu Tech) The adjustment covers V4 Pro and V4 Flash and introduces peak-valley pricing for the V4 series. Beijing … Read more
Gasgoo Embodied Intelligence has learned that Light Origins has closed a several-hundred-million-yuan Pre-A round. The round was led by Guoke Investment, with CMG Venture Capital and Xianghe Capital joining, and the funds will go to large-scale embodied-model training, multimodal data infrastructure, full-stack hardware-software platform research, and expansion of a high-density talent team. Light Origins, founded … Read more
Canchen Technology has completed a seed, angel and angel-plus series of rounds totalling tens of millions of yuan, with backers including Oriental Fortune, Songhe Capital and several university and government-linked funds. The firm builds system-level virtual-organ spatio-temporal dynamic evolution models, a full-scale simulation platform from molecule and cell through tissue to organ, using AI digital … Read more
MiFeng Technology, which runs a one-stop physical-AI data service platform, has closed a new round of several hundred million yuan. The round was led by China Telecom, with Zhangjiang Group following and Sequoia China and Yuanqi Innovation topping up their earlier stakes. The money will scale production of the MEgo series of bodiless capture products, … Read more
Yuanchuan Micro, a startup building inference chips, has closed a Pre-A round of several hundred million yuan. The round was led by IDG Capital, with Futong Capital, Jiukun Venture, Shangqi Capital, Shunxi Fund and Xuhui Science City Investment joining. It is the firm’s fourth raise since the start of 2026, and the money will go … Read more
Reuters reports that Apple is training a China-specific large language model with Alibaba, which supplies compute and localisation support. The model is trained independently for the China market rather than translated from an overseas version, a notable shift in Apple’s China AI strategy. Apple and Alibaba are reported to be co-training a China-specific model. (Source: … Read more
A Zhejiang University and Swansea University team took an unintuitive finding to IJCAI: when fine-tuning a vision-language model on new data, tuning the text tower helps and tuning the image tower can hurt. Their Adaptive Asymmetric Adapter automates that choice, applying the brakes to visual tuning exactly when it would damage the model. The adaptive … Read more
At CVPR 2026 the 3D Gaussian Splatting field stopped chasing prettier renders and started chasing industry. Across eight representative papers, the work closes the pipeline from feed-forward, second-scale generation to GPU-operator rebuilds, noise-robust SLAM and Neural ODE physics, turning 3DGS from a 3D photo album into the physical base of embodied AI and world models. … Read more
Tencent’s second-quarter results show new AI products cut 10.5 billion yuan from operating profit, while capital spending hit 52.78 billion yuan, up 176 per cent, as it bets on WorkBuddy, the WeChat a
Paul Meade, the Apple veteran who led Vision Pro hardware for seven years, is joining OpenAI’s hardware team after a reorganisation under incoming CEO John Ternus, a sign OpenAI is moving from concept