Moonshot AI
Beijing, China AI company that makes the Kimi AI chat service
Profile
- Country
- China
- Headquarters
- Beijing, China
- Founded
- 2023
- Founders
- Yang Zhilin (杨植麟) with fellow Tsinghua alumni
- Type
- Private, unlisted company as of October 2026 (investors include Alibaba and Tencent)
Moonshot AI is an AI company headquartered in Beijing, China. Its Chinese name, Yuezhi Anmian (月之暗面), means 'the dark side of the moon'. It was founded in 2023 by Yang Zhilin, who graduated from Tsinghua University and earned a PhD at Carnegie Mellon University in the US, together with fellow Tsinghua alumni. As of October 2026, Yang Zhilin is the CEO, and the company is a private company that is not yet listed on a stock exchange. Large Chinese tech companies such as Alibaba and Tencent have invested in it, and in October 2026 Bloomberg and other media reported that it was preparing to list on the Hong Kong stock exchange.
Its main service is Kimi, an AI chat service first launched in October 2023. Kimi can be used through a smartphone app, on the web and as a desktop program. The company has also released tools such as Kimi Code, which helps with programming. For developers, it offers an API (a channel through which other programs can call an AI model) and charges according to use, and it also sells business plans for companies and organisations. The company says its research is aimed at artificial general intelligence (AGI, AI that can do many different tasks the way a person can).
Moonshot AI has released large models as open weights (models published so that anyone can download and use them). Kimi K2, released in July 2025, is a 'mixture-of-experts' (MoE) model with one trillion parameters (the adjustable values inside a model); for each token (small piece of text) it processes, it uses only 32 billion of them. It was published under a 'Modified MIT License', whose only added condition is that commercial services with more than 100 million monthly users or more than 20 million US dollars in monthly revenue must display 'Kimi K2' on screen. Kimi K3, announced in July 2026, has 2.8 trillion parameters, understands text and images together, and can read up to one million tokens at once. K3 uses its own 'Kimi K3 License': a company that runs a business offering AI models to outside users (for example through an API), and whose revenue together with its affiliates exceeds 20 million US dollars over any 12 consecutive months, must sign a separate agreement with Moonshot AI before using it commercially, and very large commercial services must also display 'Kimi K3'.
Sources for this profile (13) · as of Nov 1, 2026
- Moonshot AI (official site) · Moonshot AI
- Kimi K3: Open Frontier Intelligence · Moonshot AI (Kimi blog)
- moonshotai/Kimi-K3 model card · Hugging Face (Moonshot AI)
- Kimi K3 License (LICENSE file) · Hugging Face (Moonshot AI)
- moonshotai/Kimi-K2-Instruct model card and Modified MIT License · Hugging Face (Moonshot AI)
- Kimi API Platform · Moonshot AI
- Chinese start-up Moonshot AI raises US$1 billion in funding round led by Alibaba and VC HongShan · South China Morning Post
- Who Is Moonshot AI? · The Wire China
- Moonshot hits AI mark with Kimi K3 launch (Reuters) · RTHK / Reuters
- Moonshot AI · Wikipedia
- Kimi K2 LICENSE (Modified MIT License) · GitHub (Moonshot AI)
- Meet Yang Zhilin, the CEO and founder behind China's buzzy new Kimi K3 AI model · Business Insider
- Moonshot said to eye Hong Kong IPO in early 2027 after value hits US$50 billion · The Business Times (Bloomberg)
AIMade by AIThis profile was written by AI (Claude) from the 13 sources above (official pages, filings, news reports).How it's made
Records 14
Newest first. Open one to see it with its source.
An open-weight model in the 3-trillion-parameter class. A prime example of the pattern in which scale grows while licenses narrow
The person involved makes explicit that open-weight decisions are a function of competitive ranking, not ideology. The most direct explanation yet of the motive behind China's open-source strategy
An open-weight coding model that puts token efficiency up front
Raised $3.9 billion over 6 months. A case showing that open-weight distribution works commercially
An open-weight model entered the 80% range on SWE-Bench Verified. No independent party checking the result is named
The point where the K2 line shifted from text to native multimodality, and the arrival of an open-weight agent swarm
A statement in which a Chinese lab justified open weights in normative language, and a record of a party directly denying the low-cost narrative
An open-weight reasoning agent trained to interleave thinking with tool calls; Moonshot claims it sustains 200–300 sequential tool calls without human intervention
The first to beat full attention in a fair comparison; up to 75% less KV cache and up to 6× faster decoding at 1M context
The first 1T-class open-weight model, and a new type of license with revenue and scale thresholds
Reasoning models were immediately reproduced as open weights, and 2 Chinese labs declared their pursuit of the frontier on the same day
Tiered KV cache management grounded in a production service
The first systematic statement by a Chinese lab founder declaring long context a core pillar of the path to AGI. Supplied the language for the 2024 reshaping of China's LLM race into a long-context race