Moonshot AI
Moonshot AI is an artificial intelligence company focused on developing large language models to achieve AGI (Artificial General Intelligence). The company develops foundation models including the Kimi chatbot and open-source LLMs with long context capabilities.
At a Glance
- Developers building AI applications
- Enterprises building multi-modal AI agents
- International teams and businesses
- Content creators
- +6 more
AI Tools by Moonshot AI
(5)Kimi
Multimodal AI Assistant with Agents
Kimi Claw
Cloud Hosted AI Agent Platform
Kimi Code
AI Coding Assistant
Kimi K2
Open Source Multimodal Agent Framework
Kimi AI
AI Assistant with Long Context
Discussions
No discussions yet
Be the first to start a discussion about Moonshot AI
Latest News
Moonshot AI releases Kimi K2.5 multimodal model and Kimi Code coding agent, most powerful open-source LLM
Moonshot AI valuation increases to $4.8 billion in new funding round
IDG leads $500M Series C for Moonshot AI at $4.3 billion valuation, oversubscribed by Alibaba, Tencent
Moonshot AI releases Kimi K2 Thinking model for advanced reasoning, outperforming GPT-5 and Claude Sonnet 4.5 in benchmarks
Products & Services
ChatGPT-like chatbot built on Moonshot's proprietary LLM with capabilities in handling extremely long text queries of up to 2 million Chinese characters. Surged in popularity in China months before DeepSeek's release. Supports online search, deep thinking, multimodal reasoning, and long-form text conversations.
China's first open-source trillion-parameter large language model with mixture-of-experts (MoE) architecture. Has 1 trillion total parameters but activates 32 billion parameters at a time. Designed for code generation and agentic problem-solving with 128,000 token context window. Uses MuonClip optimizer to prevent training instability. Scored 65.8% on SWE-Bench Verified, 53.7% on LiveCodeBench, and 97.4% on MATH-500.
Open-source update to Kimi K2 designed for advanced reasoning and agentic tasks. Outperformed OpenAI's GPT-5 and Anthropic's Claude Sonnet 4.5 in several benchmarks according to reports.
Natively multimodal open-source model that understands text, image, and video inputs. Trained on 15 trillion mixed visual and text tokens. Features agent swarm orchestration that can self-direct up to 100 sub-agents and execute parallel workflows of up to 1,500 tool calls. Includes autonomous visual debugging to fix layout shifts or aesthetic errors. Scored 50.2% on Humanity's Last Exam (HLE) and 76.8% on SWE-bench Verified. Beats GPT 5.2 and Gemini 3 Pro on SWE-Bench Multilingual, and outperforms GPT 5.2 and Claude Opus 4.5 on VideoMMMU. Released under Modified MIT License.
Market Position
Moonshot AI is positioned as one of China's leading 'AI Tiger' companies and a key competitor to OpenAI, DeepSeek, Alibaba's Qwen, and other Chinese AI firms. The Kimi chatbot was described as 'China's answer to ChatGPT' and surged in popularity before DeepSeek's release. Kimi K2.5 narrows the US-China AI model development gap to its closest point ever. The company differentiates on long context windows (up to 2M characters) combined with RAG UX, and has been praised for prioritizing consumer experience. Recognized in a US government report for demonstrating the 'growing depth' of China's AI industry. Competes with OpenAI, Anthropic, Google, DeepSeek, Alibaba (Qwen), Zhipu, MiniMax, and Z.AI.
Leadership
Founders
Yang Zhilin
Born 1992, earned his PhD at Carnegie Mellon University after graduating first in his class from Tsinghua University's Computer Science Department, completing PhD in under 4 years. Published over 20 influential papers in AI, collaborated with multiple Turing Award winners including Meta chief AI scientist Yann LeCun, Google research scientist Quoc V Le, and Yoshua Bengio. Worked at Facebook AI Research (Meta) and Google Brain during and after doctoral study. In 2020, worked with Huawei Technologies on Pangu AI model. In 2021, led team on Wudao LLM at Beijing Academy of AI. Was drummer in a band called Skip List. Founded Moonshot AI in March 2023 after trip to US in December 2022.
Zhou Xinyu
Graduated with Bachelor of Science in Computer Science and Technology from Tsinghua University (2015). Worked as research intern at Tencent where he gained experience with large-scale machine learning and AI system deployment. Co-creator of ShuffleNet, a mobile-friendly CNN architecture that became instrumental in edge AI and vision applications, cited and deployed by industry giants like Apple.
Wu Yuxin
Obtained undergraduate degree in Computer Science from Tsinghua University (2015), then received master's in Computer Vision from Carnegie Mellon University. Worked at Google Brain on foundation model research and at Facebook AI Research (FAIR) with major contributions to computer vision and AI infrastructure systems. Creator of Detectron2, an open-source computer vision library widely adopted in research and production environments. His open-source projects support training and deployment for dozens of products within Meta and other industry leaders.
Executive Team
Yang Zhilin
CEO and Founder
31-year-old (as of 2026) computer scientist with PhD from Carnegie Mellon University, ranked first in Tsinghua University's Computer Science Department. Published 20+ influential AI papers, worked at Meta AI and Google Brain. Previously involved in Huawei's Pangu AI and Beijing Academy of AI's Wudao LLM.
Founding Story
Yang Zhilin co-founded Moonshot AI in March 2023 after a trip to the US in December 2022 to learn about GenAI applications and recruit talent for his AI venture. The company name 'Moonshot' comes from Pink Floyd's album 'Dark Side of the Moon.' Yang explained: 'For us, it's about exploring the unknown. Just like AGI, you usually only see the illuminated side of the moon, but the dark side remains mysterious. It's challenging, yet full of potential. That aligns with our mission.' Yang and his team describe themselves as 'stubborn AGI purists' who believe 'AGI is the only thing worth doing in the next decade' and that 'vision matters most.' Their vision was to develop a company that combines 'OpenAI's technology idealism with the business philosophy of ByteDance.' The three founders were schoolmates at Tsinghua University.
Business Model
Revenue Model
API usage (pay-per-token pricing), subscription plans for consumer chatbot (freemium to premium conversion), model-as-a-service deals, and API licensing. Open-source models under Modified MIT License with attribution requirements for entities with 100+ million monthly users or $20 million+ monthly revenue.
Pricing Tiers
Four days of priority use of Kimi chatbot
One year of priority use of Kimi chatbot (6 total tiers ranging from 5.2 yuan for 4 days to 399 yuan annually)
Standard input pricing for multimodal model, context window 262,144 tokens
Cached input pricing for multimodal model
Output pricing for multimodal model
For K2 generation models
Cached input for K2 generation models
Output pricing for K2 generation models
Context window 131,072 tokens
Context window 131,072 tokens
Target Markets
- Developers building AI applications
- Enterprises building multi-modal AI agents
- International teams and businesses
- Content creators
- Researchers and academic institutions
- China's tech ecosystem (powering third-party apps, mini-programs, enterprise deployments)
- Long document analysis and processing
- Coding and software development
- Agent-based task automation
- Visual analysis applications
- Multi-modal AI agents for scenario-specific automation
- Enterprise RAG (Retrieval-Augmented Generation) applications