China AI Weekly Digest summarizes the weekās big news, model updates, policy, products, and research, sourced from the aggregated newsfeed of China AI Index. The digest below is mostly compiled and drafted by an AI agent sourcing from credible news outlets, then fact-checked and edited by me.
This week: 322 stories tracked (5 editor picks, 112 English-language, 205 Chinese-language šØš³) over 2026-08-22ā2026-08-29.
The Big Three
A mystery model that quietly topped the charts turns out to be Zhipuās, running entirely on domestic Chinese chips. An anonymous model calling itself āOx Alphaā had developers guessing for six days. Some speculated it was Cursor trained on open GLM weights. Then Zhipu (Z.ai) claimed it as GLM-5.3-Flash: a natively multimodal, 320-billion-parameter (18B active) model running inference entirely on domestic Chinese chips, priced at roughly 1/40 of Claude Opus 4.8. Z.aiās shares jumped on the reveal, and two days later the company open-weighted its larger flagship, GLM-5.3 (744B total/40B active), under an MIT license. (SCMP/Bloomberg/CNBC/@Zai_org)
DeepSeek edges toward a 2027 listing as revenue rockets and its valuation nears $74 billion. DeepSeek is closing in on a pre-IPO funding round as it lines up a public-market debut as early as 2027, with the Wall Street Journal reporting a target valuation of $74 billion; The Information separately reported DeepSeekās revenue reached $70 million in July, a tenfold jump from a year earlier. The fundraising push doubles as a coming-out moment for founder Liang Wenfengās quant fund High-Flyer, which bankrolled DeepSeek from the start and is now navigating Chinaās crowded, choppy IPO market on its behalf. (WSJ/CNBC/SCMP/The Information)
A robot beats Usain Boltās 100m record at Chinaās World Humanoid Robot Games. Beijing hosted the second edition of the World Humanoid Robot Games, a five-day, 51-event spectacle spanning running, table tennis, soccer and boxing. The event has become a showcase for Chinaās robotics progress. This yearās edition featured a robot beating Usain Boltās own 100m sprint record. Homegrown humanoid maker AGIBOT dominated the games, topping both the gold-medal and overall-medal tables with 46 medals in total (18 gold, 16 silver, 12 bronze). (AP/BBC/@AGIBOTofficial)
Models
A wave of new releases crowded the weekās model news.
Tencentās Hunyuan team previewed Hy4, featuring 770B parameters, 49B active, a 1M-token context window. Tencent says the model outperforms both Z.ai and Moonshot. (Bloomberg/@TencentHunyuan)
Alibabaās Qwen team launched Qwen3.8-Flash, a smaller multimodal MoE model billed as an early preview of the Qwen4 architecture, cutting training and serving costs. (Bloomberg/@Alibaba_Qwen)
MiniMax said it has begun merging its M3 and H3 models into a single multimodal model and expanded its Alibaba Cloud compute partnership to keep pace with training and inference demand. (CNBC/SCMP)
Xiaomi is developing a new self-designed āXringā chip and has lined up TSMC to manufacture it, sources told Reuters. (Reuters)
Ant Groupās Ling team shipped Ling-3.0-flash-Fin, built for financial-sector use with financial institutions and domain experts. (@AntLingAGI)
Funding
Money kept moving fast this week into AI labs, and out of some newly public robotics stocks.
Alibaba completed a HK$80 billion (~$10.2 billion) share placement that was nearly 3x oversubscribed, with more than 40% of orders from sovereign wealth and other long-term funds, and said proceeds will go toward AI infrastructure, chips, models and applications. Alibabaās own top executives followed up by buying roughly $15 million of shares themselves. (Reuters/FT/Bloomberg/SCMP)
Moonshot AI is preparing a Hong Kong IPO within six months at a valuation as high as $50 billion, Nikkei reported, after its open-weight Kimi K3 model triggered a fresh āDeepSeek momentā in global markets; Reuters separately reported Moonshot is in talks with Microsoft, Amazon and Google to host K3 on their clouds for up to a 30% revenue share, which would be a first for a Chinese AI lab. (Nikkei Asia/Reuters)
MiniMaxās annualized revenue jumped to $800 million, The Information reported, while Chinese media separately reported the surge at roughly 500% ARR growth and 2,000% token growth, with first-half revenue already exceeding all of 2025. (The Information/éåä½)
Xpengās robotics unit raised over $900 million from investors including Alibaba, valuing the unit near $6.3 billion, even as Xpengās own quarterly earnings and delivery forecast disappointed investors. (Bloomberg/CNBC/SCMP)
Newly listed Unitreeās stock slumped hard enough to stoke bubble fears across Chinese humanoid robotics, while chipmaker CXMTās stock rallied 580% ahead of its first earnings report since going public. (SCMP/Bloomberg)
AI chipmaker Enflame set a subscription date for its roughly $900 million Shanghai STAR Market IPO. (Reuters)
SenseTime reported its first-ever IFRS profit since listing: RMB 2.91 billion in the first half of 2026. (@SenseTime_AI)
Policy
The Guardian reports Chinaās government has grown wary that AI companion apps could substitute for human intimacy, even as talking to AI has become mainstream there. (Guardian)
Brazilās data-protection authority fined ByteDance over unlawful processing of teenagersā data on TikTok. (Reuters/č“¢ę°)
Products
ByteDance is folding its AI tools, TRAE and Coze (ę£å), into Doubao, building a unified āDoubao Workā brand to take on Tencentās Workbuddy and Alibabaās Qwen Work. (Bloomberg/36Kr)
DeepSeekās family of cheap, open-weight models is leading a broader surge of Chinese models onto a major US model-hosting platform. (SCMP)
Baiduās ERNIE-powered assistant rolled out upgrades at Baidu AI Day, featuring 400+ professional skills across 30+ industries, with daily active users up 83% year-over-year in June. (@Baidu_Inc)
Research
Security researchers say Chinese hackers are using DeepSeek to help scale their attacks, Bloomberg reports. (Bloomberg)
Meituanās LongCat team evaluated seven frontier models on 36 AI R&D tasks to test whether agents can act as autonomous researchers. (@Meituan_LongCat)