News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM

Open-model community, China

Hangzhou

Alibaba's Qwen and DeepSeek: the city behind two of the world's most used open-weights families.

Hangzhou, capital of Zhejiang province, is home to two of the most influential open-weights model makers. Alibaba, headquartered here, builds the Qwen family through Alibaba Cloud, and DeepSeek, founded in 2023 and funded by the local hedge fund High-Flyer, surprised the world with DeepSeek-V3 and R1. Zhejiang University supplies many of their researchers.

Qwen

Alibaba launched Tongyi Qianwen (Qwen) in beta in April 2023 and has released open weights in many sizes since, mostly under Apache 2.0. Qwen3 (April 2025) included dense and mixture-of-experts models trained on 36 trillion tokens covering 119 languages and dialects. Qwen models are among the most downloaded open models and a common base for fine-tunes worldwide, including K2 Think in Abu Dhabi.

DeepSeek

DeepSeek, led by Liang Wenfeng, released V3 in December 2024, a 671-billion-parameter mixture of experts whose final training run its paper put at about 2.8 million GPU hours, and R1 in January 2025 under the MIT licence. R1's app briefly topped the US App Store, and several governments restricted it on official devices over data concerns. Local media grouped DeepSeek with five other start-ups as Hangzhou's six little dragons.

Models of Hangzhou

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.