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

πŸ‡¨πŸ‡³ Models of China

100 models from China, best known first.

China has pursued a state-led AI strategy since 2016, aiming for global leadership by 2030. The sector blends private innovation with strict government oversight and military integration. Chinese models prioritize ideological alignment and data security within domestic regulations.

  • State Council aims for global AI leadership by 2030
  • Fifteen companies designated as national AI teams
  • Data Security Law published in 2021
  • Baidu holds largest LLM market share at 19.9%
  • World Artificial Intelligence Cooperation Organization created in 2026

Language models in China

China stands as a top global player in artificial intelligence, driven by coordinated state planning and rapid private sector growth. The government views the technology as essential for economic transformation and national security. Policies encourage collaboration between universities, enterprises, and the military to accelerate development. This approach has created a robust ecosystem of large language models and specialized applications across various industries.

History

Research began in the late 1970s following Deng Xiaoping's reforms. Early efforts focused on cybernetics and academic journals. The field gained momentum in the 1980s with institutional support and international conferences. A major shift occurred in 2017 when the State Council issued a strategic plan targeting global dominance by 2030. Subsequent years saw increased funding and the launch of foundational large language models by research institutes.

Labs and companies

The state designates fifteen national teams, including Baidu, Tencent, Alibaba, SenseTime, and iFlytek. These firms lead specific sectors like facial recognition and speech processing. By 2024, Baidu held nearly twenty percent of the large language model market. Other key players include Zhipu AI, Baichuan, Moonshot AI, and MiniMax. Huawei also released significant versions of its Pangu model during this period.

Languages

Chinese models are trained primarily on local data, reflecting government priorities and Core Socialist Values. Regulations require content to avoid discrimination and respect intellectual property. Models often express views aligned with the ruling party. Some systems have been explicitly trained on political thought documents. The industry relies heavily on open-source variants to achieve widespread adoption globally.

Policy and programmes

China enforces strict rules on data security and algorithm transparency. Laws mandate that generative AI services uphold ideological standards and safeguard user privacy. Companies must test models regularly and minimize unsafe training data. Export controls from the US have limited access to advanced chips. Recent measures restrict state use of certain foreign tools and promote domestic alternatives through initiatives like AI+.

Today

In 2026, China launched the World Artificial Intelligence Cooperation Organization. Government restrictions continue to shape model deployment and data usage. Domestic firms compete fiercely while navigating regulatory requirements. International engagement includes diplomatic efforts to influence global governance norms. The sector remains focused on integrating AI into the real economy and maintaining technological self-sufficiency.

Where this page comes from

The facts in this overview come from the Wikipedia article "Artificial intelligence industry in China", written by its contributors and used under the CC BY-SA 4.0 licence. EveryLLM rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

AI lab cities in China

fnLongCat-2.0 πŸ‡¨πŸ‡³ ChinaLongCat-2.0 A 1.6 trillion parameter MoE language model from Meituan.General Β· Open weights MiniCPM πŸ‡¨πŸ‡³ ChinaMiniCPM Efficient small language models for end-side devicesGeneral Β· Open weights Β· 2024 fnGLM-5.3 πŸ‡¨πŸ‡³ ChinaGLM-5.3 Open-weights coding and cyber-security model from Z.aiCode Β· Open weights Β· 2026 imgCogVLM πŸ‡¨πŸ‡³ ChinaCogVLM Open-source visual language model with deep feature fusionMultimodal Β· Open weights Β· 2023 MiniMax-Text-01 πŸ‡¨πŸ‡³ ChinaMiniMax-Text-01 A large language model with a four million token context window.General Β· Open weights Β· 2025 fnMiniMax-M1 πŸ‡¨πŸ‡³ ChinaMiniMax-M1 Open-weight reasoning model with hybrid attention and long contextReasoning Β· Open weights Β· 2025 fnLongCat-Flash πŸ‡¨πŸ‡³ ChinaLongCat-Flash Efficient 560B parameter MoE model for agentic tasksGeneral Β· Open weights Β· 2025 PanGu-Sigma πŸ‡¨πŸ‡³ ChinaPanGu-Sigma Trillion-parameter sparse language model for Chinese NLP tasksGeneral Β· Research only Β· 2023 fnGLM-5 πŸ‡¨πŸ‡³ ChinaGLM-5 Open-weight model for engineering and agentic tasksGeneral Β· Open weights Β· 2026 fnMiniMax-M2.5 πŸ‡¨πŸ‡³ ChinaMiniMax-M2.5 Cost-efficient coding and agentic model from MiniMax AICode Β· API only Β· 2026 ERNIE 3.0 πŸ‡¨πŸ‡³ ChinaERNIE 3.0 Knowledge-enhanced pre-training model for language tasksGeneral Β· Research only Β· 2021 ERNIE 3.0 Titan πŸ‡¨πŸ‡³ ChinaERNIE 3.0 Titan Large-scale Chinese knowledge-enhanced pre-trained language modelGeneral Β· Research only Β· 2021 ERNIE 1.0 πŸ‡¨πŸ‡³ ChinaERNIE 1.0 Knowledge-enhanced language model for Chinese NLP tasks.Encoder Β· Research only Β· 2019 fnERNIE-4.5-21B-A3B-Thinking πŸ‡¨πŸ‡³ ChinaERNIE-4.5-21B-A3B-Thinking Lightweight reasoning model with 21 billion parametersReasoning Β· Open weights Β· 2025 imgERNIE 4.5 πŸ‡¨πŸ‡³ ChinaERNIE 4.5 Baidu's multimodal mixture-of-experts language model series.General Β· Open weights Β· 2025 fnHunyuan-A13B πŸ‡¨πŸ‡³ ChinaHunyuan-A13B Open-source MoE language model with 80 billion total parametersGeneral Β· Open weights Β· 2025 Hunyuan-TurboS πŸ‡¨πŸ‡³ ChinaHunyuan-TurboS Hybrid Mamba-Transformer MoE model by TencentGeneral Β· API only Β· 2025 Seed1.5-Thinking πŸ‡¨πŸ‡³ ChinaSeed1.5-Thinking A reasoning model from ByteDance Seed using reinforcement learningReasoning Β· API only Β· 2025 imgStep3 πŸ‡¨πŸ‡³ ChinaStep3 Multimodal reasoning model with 321 billion parametersMultimodal Β· Open weights Β· 2025 fnSeed-OSS-36B πŸ‡¨πŸ‡³ ChinaSeed-OSS-36B Open-source 36B parameter language model with controllable reasoning.General Β· Open weights Β· 2025 imgvidIntern-S1 πŸ‡¨πŸ‡³ ChinaIntern-S1 Open-source multimodal reasoning model for scienceReasoning Β· Open weights Β· 2025 MiMo-7B πŸ‡¨πŸ‡³ ChinaMiMo-7B Open reasoning model from Xiaomi for maths and codeReasoning Β· Open weights Β· 2025 Pangu Ultra πŸ‡¨πŸ‡³ ChinaPangu Ultra A 135-billion parameter dense language model trained on Ascend hardware.General Β· API only Β· 2025 imgvidfnInternVL3 πŸ‡¨πŸ‡³ ChinaInternVL3 Open-source multimodal large language model seriesMultimodal Β· Open weights Β· 2025 imgfnUI-TARS πŸ‡¨πŸ‡³ ChinaUI-TARS Open-source multimodal agent for GUI tasksMultimodal Β· Open weights Β· 2025 MiMo-V2.5-Pro πŸ‡¨πŸ‡³ ChinaMiMo-V2.5-Pro Open-source MoE language model with 1M token contextGeneral Β· Open weights Baichuan-M2 πŸ‡¨πŸ‡³ ChinaBaichuan-M2 Open-source medical reasoning model with 32 billion parameters.Reasoning Β· Open weights Β· 2025 Baichuan 2 πŸ‡¨πŸ‡³ ChinaBaichuan 2 Open multilingual language models with 7B and 13B parameters.General Β· Open weights Β· 2023 imgwavvidMiMo-V2.6-Pro πŸ‡¨πŸ‡³ ChinaMiMo-V2.6-Pro Xiaomi’s 1T parameter multimodal model with 1M token contextMultimodal Β· Open weights fnStep-3.5-Flash πŸ‡¨πŸ‡³ ChinaStep-3.5-Flash Open-source MoE model for reasoning and agentsGeneral Β· Open weights Β· 2026 Qwen2.5-72B πŸ‡¨πŸ‡³ ChinaQwen2.5-72B A 72-billion parameter open-weight language model from Alibaba Cloud.General Β· Open weights Β· 2024 DeepSeek Coder πŸ‡¨πŸ‡³ ChinaDeepSeek Coder Open-source code generation models from DeepSeekCode Β· Open weights Β· 2024 Qwen2 πŸ‡¨πŸ‡³ ChinaQwen2 Open-weight language model series from AlibabaGeneral Β· Open weights Β· 2024 imgvidGLM-4.1V-Thinking πŸ‡¨πŸ‡³ ChinaGLM-4.1V-Thinking Open-source vision-language model with reasoning focusMultimodal Β· Open weights Β· 2025 Qwen2.5-Coder πŸ‡¨πŸ‡³ ChinaQwen2.5-Coder Open code-specialised language model series by Alibaba CloudCode Β· Open weights Β· 2024 imgvidfnKimi K2.5 πŸ‡¨πŸ‡³ ChinaKimi K2.5 Open-source multimodal agentic model with vision and reasoning.Multimodal Β· Open weights Hunyuan-Large πŸ‡¨πŸ‡³ ChinaHunyuan-Large Open-source mixture of experts model with 389 billion parametersGeneral Β· Open weights Β· 2024 fnMiniMax-M2 πŸ‡¨πŸ‡³ ChinaMiniMax-M2 Open-source MoE model for coding and agentsCode Β· Open weights imgvidfnKimi K3 πŸ‡¨πŸ‡³ ChinaKimi K3 Open-weight multimodal agentic model with 2.8T parametersGeneral Β· Open weights Β· 2026 imgwavvidQwen2.5-Omni πŸ‡¨πŸ‡³ ChinaQwen2.5-Omni End-to-end multimodal model for real-time speech and video chat.Multimodal Β· Open weights Β· 2025 fnInternLM2 πŸ‡¨πŸ‡³ ChinaInternLM2 Open-source 20 billion parameter language model with long context supportGeneral Β· Open weights Β· 2024 fnMiniMax-M2.7 πŸ‡¨πŸ‡³ ChinaMiniMax-M2.7 Open-weight model for coding and agent tasksCode Β· Open weights DeepSeek-V2 πŸ‡¨πŸ‡³ ChinaDeepSeek-V2 Efficient mixture-of-experts language model with 236 billion parametersGeneral Β· Open weights Β· 2024 DeepSeek-V3 πŸ‡¨πŸ‡³ ChinaDeepSeek-V3 Open-weight mixture of experts language model with 671 billion parameters.General Β· Open weights Β· 2024 imgwavvidQwen3-Omni πŸ‡¨πŸ‡³ ChinaQwen3-Omni Multimodal model processing text, audio, video and images with real-time speech output.Multimodal Β· Open weights Β· 2025 fnKimi K2 Thinking πŸ‡¨πŸ‡³ ChinaKimi K2 Thinking Open-source reasoning model with tool use and INT4 quantizationReasoning Β· Open weights Qwen2.5-Max πŸ‡¨πŸ‡³ ChinaQwen2.5-Max Large-scale mixture of experts model by Alibaba CloudGeneral Β· API only Β· 2025 fnGLM-4.6 πŸ‡¨πŸ‡³ ChinaGLM-4.6 A large language model with a 200K token context window.General Β· API only

The weekly digest

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