2026-03-09
OpenClaw adoption in China is accelerating across all the platform. Wuxi become the 2nd govt to actively encourage its adoption w/ 12 new policies + subsidies. ByteDance rolled out ArkClaw for 9.9 RMB or 49.9 RMB/month supporting Doubao, Kimi, Minimax & GLM. Tencent cloud is [image]
Bloomberg
OpenClaw mania hits China: AI labs launch tools to help users deploy OpenClaw, a Shenzhen district drafts a policy encouraging free OpenClaw services, and more
Chinese software shares surged Monday after local government agencies joined tech leaders such as Tencent Holdings Ltd …
2026-03-03
Interesting report here from Minimax as the 1st to report among AI shops. $79m revenue & 20m gross profit. Gross margin improved from 12 to 25% in 2025. Their R&D expenses was $252.8m I don't know what is “Fair Value Loss on Financial Liabilities”, but it's said to be “related [image]
Bloomberg
Shanghai-based MiniMax reports 2025 revenue up 159% YoY to $79M, above $71.4M est., and a $1.87B net loss, up from a $465.2M net loss in 2024, after its IPO
2026-03-02
Interesting report here from Minimax as the 1st to report among AI shops. $79m revenue & 20m gross profit. Gross margin improved from 12 to 25% in 2025. Their R&D expenses was $252.8m I don't know what is “Fair Value Loss on Financial Liabilities”, but it's said to be “related [image]
Bloomberg
Shanghai-based MiniMax reports 2025 revenue up 159% YoY to $79M, above $71.4M est., and a $1.87B net loss, up from a $465.2M net loss in 2024, after its IPO
MiniMax Group Inc. reported a better-than-expected 159% surge in revenue in 2025, reflecting the torrid growth that's drawn investors to China's leading OpenAI rivals.
2026-02-17
Qwen 3.5 has been released. 397B parameters including 17B active ones. This is the flagship open sourced version. Looks to be reasonably better than Qwen3 235B. I've found larger Qwen models to not be all that great. Given the recent Minimax & GLM releases, Qwen has to also have [image]
Reuters
Alibaba debuts Qwen3.5, a 397B-parameter open-weight multimodal AI model that it says is 60% cheaper to use and 8x better at large workloads than Qwen3
2026-02-16
Qwen 3.5 has been released. 397B parameters including 17B active ones. This is the flagship open sourced version. Looks to be reasonably better than Qwen3 235B. I've found larger Qwen models to not be all that great. Given the recent Minimax & GLM releases, Qwen has to also have [image]
Reuters
Alibaba debuts Qwen3.5, a 397B-parameter open-weight multimodal AI model that it says is 60% cheaper to use and 8x better at large workloads than Qwen3
2026-02-10
Seedance 2.0 is blowing up on my X timeline. I have not seen any video models that have blown up to this degree. Also the most attention paid to a Chinese model since DeepSeek R1 came out. I can see Seedream 5.0 on CapCut now, but Seedance 2.0 is not there yet. As long as
The Information
ByteDance debuts Seedance 2.0, a new version of its AI video generation model that can produce multi-shot scenes, available to select Jimeng and Jianying users
2026-02-09
Seedance 2.0 is blowing up on my X timeline. I have not seen any video models that have blown up to this degree. Also the most attention paid to a Chinese model since DeepSeek R1 came out. I can see Seedream 5.0 on CapCut now, but Seedance 2.0 is not there yet. As long as
The Information
ByteDance debuts Seedance 2.0, a new version of its AI video generation model that can produce multi-shot scenes, available to select Jimeng and Jianying users
ByteDance's new AI video generation model Seedance 2.0, unveiled over the weekend, is generating a lot of buzz on social media, with some calling it a game changer.
2025-12-23
Comparing MiniMax M2.1 to GLM-4.7, it seems to do really well in coding related stuff but still trails Zai in reasoning & browsing related stuff. Either way, pretty good effort here by Minimax. Chinese open src models still trail US hyperscaler's frontier model, but the intense [image]
MiniMax
China's MiniMax releases M2.1, an upgrade to its open-source M2 model that it says has “significantly enhanced” coding capabilities in Rust, Java, and others
MiniMax has been continuously transforming itself in a more AI-native way. The core driving forces of this process are models …
2025-11-07
It is really a good thing when open source AI lab stops just comparing themselves to other open src models, but compare to the best closed src ones. I will have to test out Kimi K2, but this looks very promising.
CNBC
Chinese startup Moonshot releases Kimi K2 Thinking, an open-weight model it claims beats GPT-5 in agentic capabilities; source: the model cost $4.6M to train
Chinese startup Moonshot on Thursday released its latest generative artificial intelligence model which claims to beat OpenAI's ChatGPT in …
2025-08-27
SMIC's advanced node are in its SMSC subsidiary, announced in 2016, w/ plan for 35k wpm in phase 1 (SN1) & another 35k in phase2 (SN2) Last yr, they already imported necessary ASML scanners for full 70k. Ramping up capacity required validation of US-free supply chain for 5-7nm. [image]
Financial Times
Sources: a Chinese fab making Huawei's AI chips is set to start production as soon as the end of 2025, with two more due in 2026; the fabs' ownership is unclear
Fabrication plants serving Huawei push to increase production as US cuts off access to Nvidia's top processors
2025-08-19
Xiaomi Q2 earnings. Revenue up 30% YoY for the 5th quarter in a roll! Delivered 81302 EVs in Q2 for 21.3B RMB in revenue Auto division has been a huge driver of Xiaomi success in the past yr. [image]
Bloomberg
Xiaomi reports Q2 revenue up 31% YoY to ~$16.2B, above ~$16B est., and net income up 134% YoY to ~$1.6B, driven by EV sales countering slowing smartphone demand
Linda Lew / Bloomberg :
2025-05-20
Malaysia will be site of 1st HW AI chip export to oversea customers as it seeks to grow local AI industry. Last month, MYEG signed deal w/ on Malaysia-China AI Innovation & Cooperation Centre as Flagship China-ASEAN AI Lab This lab will see DeepSeek, Leju Robot, DJI, HW, Alibaba [image]
Bloomberg
Malaysia downplays its deal to use 3,000 Huawei chips to build a first-of-its-kind AI system by 2026, underscoring its delicate position in the US-China AI race
Malaysia declared it'll build a first-of-its-kind AI system powered by Huawei Technologies Co. chips, only to distance itself …
2025-03-25
Alibaba is actively hiring from elite universities like Tsinghua, Peking U, Zhejiang U, MIT & Stanford for LLMs, multi-modal, model apps & AI infrastructure ~50% of Ali's spring hiring is AI related including 80% in its cloud division & 65% in AutoNavi
Bloomberg
Alibaba Chairman Joe Tsai warns of a bubble in data center construction, saying the buildout may outpace early AI demand and is starting to look indiscriminate
A rush by big tech firms, investment funds and other entities to erect server bases from the US to Asia is starting to look indiscriminate …
Alibaba is actively hiring from elite universities like Tsinghua, Peking U, Zhejiang U, MIT & Stanford for LLMs, multi-modal, model apps & AI infrastructure ~50% of Ali's spring hiring is AI related including 80% in its cloud division & 65% in AutoNavi
Simon Willison's Weblog
Alibaba releases Qwen2.5-VL-32B, a 32B open model under Apache 2.0, claiming better math reasoning and alignment with human preferences than earlier 2.5 models
Qwen2.5-VL-32B: Smarter and Lighter. The second big open weight LLM release from China today - the first being DeepSeek v3-0324.
2025-03-23
Tencent saw revenue growth in 2024 & shelled out 70.7B RMB for R&D & 76.8B for capex Huge AI investment. Hunyuan model w/ trillion parameters, T1 reasoning + Turbo S regular model - open src Integrated HY & DS R1 in its apps for search, voice input & content generation 200m MAU [image]
South China Morning Post
Tencent unveils Hunyuan T1, a new reasoning AI model powered by its Hunyuan Turbo S AI model, and claims it rivals DeepSeek's R1 in both performance and pricing
Tencent Holdings has unveiled a new artificial intelligence (AI) reasoning model, Hunyuan T1, that rivals DeepSeek's R1 in both performance and pricing.
2025-03-12
Manus announced on Weibo it formed strategic partnership w/ Alibaba/Qwen team to utilize domestic model & computing platform to full realize all of Manus' functions Ji Yichao revealed Manu uses different fine tuned version of Qwen. Alibaba cloud will need a whole lot more compute [image]
Reuters
Manus partners with Alibaba's Qwen to integrate Manus' AI agent functions with AI models and computing platforms in China and to collaborate on Qwen's AI models
China's Manus AI announced on Tuesday a strategic partnership with the team behind tech giant Alibaba's Qwen AI models …
2025-03-02
Impressive. DeepSeek achieves maximum utilization of its H800 nodes during 24 hr period. Maximizing cache hit to deliver more for less -> can achieve theoretical margin of 545%! Even accounting for free web/app access + nighttime discount, DS still probably profitable. [image]
Bloomberg
DeepSeek says its V3 and R1 models' cost of inferencing relative to sales during a 24-hour-period on February 28 put “theoretical” profit margins at 545%
Chinese artificial intelligence phenomenon DeepSeek revealed some financial numbers on Saturday, saying its “theoretical” …
2025-03-01
Impressive. DeepSeek achieves maximum utilization of its H800 nodes during 24 hr period. Maximizing cache hit to deliver more for less -> can achieve theoretical margin of 545%! Even accounting for free web/app access + nighttime discount, DS still probably profitable. [image]
Bloomberg
DeepSeek says its V3 and R1 models' cost of inferencing relative to sales during a 24-hour-period on February 28 put “theoretical” profit margins at 545%
Chinese artificial intelligence phenomenon DeepSeek revealed some financial numbers on Saturday, saying its “theoretical” …
2025-01-28
Perplexity has now integrated R1 reasoning into their deep web searches now. Note, you can host R1 or any other open src models outside of China & make money off that. You can also deploy uncensored R1. that's what MIT license allows! The arguments against using DeepSeek models
Bloomberg
Sam Altman says DeepSeek's R1 is an “impressive model, particularly around what they're able to deliver for the price” and OpenAI “will pull up some releases”
OpenAI Chief Executive Officer Sam Altman welcomed the debut of DeepSeek's R1 model in a post on X late on Monday.
2025-01-27
Deepseek is pretty sticky. Been top the iOS store for a day now in America. Its software team is quite top notch, rolling out changes on a daily basis. Idk if DS will have the best model out of China in a yr, but their app is already super valuable after just 1 week. [image]
Bloomberg
DeepSeek's iOS app tops the App Store's Top Free Apps chart in the US, beating ChatGPT, stirring doubts in Silicon Valley about the strength of the US' AI lead
- App's lower-cost model upends premise for AI spending boom — Stocks of chip gear makers ASML and Advantest plunge