Moonshot AI: Company Overview & Kimi
Moonshot AI builds Kimi, a Chinese AI assistant with 1M token context window for document processing. Free access, API available, backed by Alibaba.
Key Takeaways
- Moonshot AI (月之暗面) is a Chinese generative AI startup founded in March 2023 in Beijing, best known for its AI assistant Kimi
- Kimi processes documents up to 1 million tokens (roughly 750,000 words) in a single session, substantially more than most competing AI assistants
- For live conversations, Kimi's context window supports 128,000 tokens (approximately 96,000 words); ChatGPT (GPT-4o) matches this figure but cannot extend to 1 million tokens for file processing
- Kimi is accessible internationally via kimi.ai, supports both Chinese and English, and offers a free tier with no upfront cost
- Developers can access the Moonshot AI API at platform.moonshot.cn, with three model variants: moonshot-v1-8k, moonshot-v1-32k, and moonshot-v1-128k
- Moonshot AI is backed by investors including Alibaba Group and HongShan (formerly Sequoia Capital China)
What Is Moonshot AI?
Moonshot AI (月之暗面, meaning "Dark Side of the Moon") is a Beijing-based Chinese generative AI startup founded in March 2023, best known for building Kimi, an AI assistant whose defining technical feature is an unusually large context window that allows it to process book-length documents in a single session.
The company sits within the generative AI category (the same broad industry as OpenAI, Anthropic, and Google DeepMind) and builds its own proprietary large language models (LLMs). In China's domestic AI market, Moonshot AI has drawn attention as an independent startup competing alongside major technology incumbents, distinguishing itself through its long-context approach rather than broad platform integration. Kimi, its flagship product, reflects that thesis: built specifically to handle text at volumes that shorter-context tools cannot accommodate without truncating earlier content.
| Field | Details |
|---|---|
| Founded | March 2023 |
| Headquarters | Beijing, China |
| Founders | Yang Zhilin (CEO), Zhang Yuhao, Zhou Xinyu |
| Key Product | Kimi (AI assistant) |
| Core Technology | Long-context large language model (LLM) |
| Context Window | Up to 128,000 tokens (conversation); up to 1 million tokens (file processing) |
| Website | kimi.ai / moonshot.cn |
| Funding Status | Series B (reported); investors include Alibaba Group, HongShan |
The Founding Story: Who Built Moonshot AI?
Moonshot AI was founded in March 2023 by Yang Zhilin (杨植麟), Zhang Yuhao, and Zhou Xinyu, three researchers with backgrounds spanning top-tier academic institutions and industry AI labs.
Yang Zhilin, who serves as co-founder and CEO, studied at Tsinghua University (清华大学) before completing graduate research at Carnegie Mellon University. While at CMU, he co-authored the XLNet (arXiv:1906.08237), a language model that advanced autoregressive pretraining approaches and achieved strong results across natural language processing benchmarks at publication. That research background matters for understanding Moonshot AI's product focus: Yang's work on language model architecture, including how models process and retain context across long sequences, informed the founding thesis that large context windows would become a critical capability gap in the AI market. Kimi is the product that thesis produced.
Zhang Yuhao and Zhou Xinyu joined Yang Zhilin as co-founders and bring complementary engineering and research backgrounds from prior industry roles. Their combined experience shaped Moonshot AI's early technical direction and product-market framing around long-context processing as a differentiated capability rather than an incidental feature.
What Is Kimi? Moonshot AI's Flagship AI Assistant
Kimi is Moonshot AI's flagship AI assistant, powered by the company's proprietary large language model (LLM) and accessible via a free web interface at kimi.ai, a mobile app, and a developer API.
The underlying technology is a large language model, the same class of AI model that powers OpenAI's ChatGPT and Anthropic's Claude. An LLM is a type of model trained on large text datasets to understand and generate human language; Moonshot AI builds and trains its own rather than licensing an existing model. Like ChatGPT and Claude, Kimi is built on transformer-based architecture, the design approach that underpins virtually all modern language models. Its core capability domain is natural language processing (NLP): understanding text inputs, generating text responses, and maintaining coherence across extended interactions.
Kimi's feature set covers the following tasks:
- Long Document Processing: Read and analyze documents up to 1 million tokens in length, then answer questions about them during file-processing sessions
- Question and Answer: Respond to factual, analytical, and open-ended questions across a wide range of topics
- Writing Assistance: Draft and revise text including articles, reports, and emails, as well as edit creative content for consistency
- Code Generation and Review: Write and explain code, debug issues, or refactor existing implementations across common programming languages
- Translation: Translate text between Chinese and English with terminological consistency maintained across long documents
- Web Search: Search for and synthesize current information from the web (availability subject to change; verify at kimi.ai)
Moonshot AI has also been developing multimodal capabilities for Kimi, including image understanding, which would place it on a similar trajectory to GPT-4V and Google Gemini. Treat this as a developing direction rather than a fully available feature at time of writing. Kimi is accessible via web, mobile, and API; the developer API is covered in the API section below.
MOONSHOT token on Bybit: Moonshot AI's growth has attracted attention from crypto traders — Bybit offers MOONSHOTUSDT perpetual futures for those interested in trading the MOONSHOT token. Bybit also covers Moonshot AI's IPO trading opportunities.
Kimi's Long Context Window: What Makes Moonshot AI Different
The context window is Moonshot AI's primary technical differentiator, and it is the feature that most clearly separates Kimi from both Western and Chinese AI assistants at the time of writing.
What Is a Context Window?
A context window is the total amount of text an AI model can read and hold in memory during a single session. Think of it as the model's working memory for a conversation: everything within that window is accessible when it generates a response, and anything outside it is gone. A short context window means the model forgets earlier parts of a long document or conversation. A large one means it can work with substantially more text without losing access to content from earlier in the session.
Moonshot AI's Context Window Sizes
Kimi's context window supports up to 128,000 tokens (where one token equals roughly three-quarters of an English word, so approximately 96,000 words, about the length of a full novel) for live conversations. For file and document processing, the context window extends to 1 million tokens, equivalent to approximately 750,000 words. To put that in concrete terms: 750,000 words is roughly equivalent to the combined text of 10 full-length novels, or a year's worth of email archives for a mid-sized organization.
Why Long Context Matters: Practical Implications
A 1 million token context window means Kimi can read an entire 300-page report in a single session and answer questions that draw from any section of the document without the model losing access to earlier pages. Tools with shorter context limits cannot do this without truncating content from the start of the document as the session progresses.
Two scenarios illustrate why this matters in practice. First: a financial analyst who uploads a 250-page due diligence report and asks Kimi to compare figures from the executive summary against data tables in the appendix gets a single coherent response, because both sections remain in context simultaneously. Second: a software engineer who pastes an entire codebase into a session can ask about relationships between files uploaded early and files added later, without re-uploading or re-establishing context between questions.
For comparison, OpenAI's ChatGPT (GPT-4o) supports a context window of up to 128,000 tokens for both conversation and file processing, and Anthropic's Claude supports up to 200,000 tokens. Kimi's 1 million token capacity for document processing exceeds both. The full side-by-side comparison appears in the comparison table below.
What Can You Use Kimi For? Key Use Cases
Kimi is used across a range of professional and personal tasks, with its long-context capabilities being particularly suited to document analysis, research synthesis, and software development. The following scenarios illustrate how the tool works in practice.
Long Document Analysis
A financial analyst uploading a 250-page due diligence report can ask Kimi questions that reference data from chapter two and chapter eighteen in the same query, because the entire document remains in context throughout the session. For readers asking whether Kimi can read long PDFs: yes, Kimi processes uploaded PDF documents using its large context window, reading the full file rather than a truncated excerpt. A shorter-context tool would begin dropping the earliest pages from memory as the document grows, producing incomplete answers that ignore content from the document's opening sections.
Research Synthesis and Summarization
A graduate researcher who uploads six academic papers on the same topic can ask Kimi to identify where the papers agree, flag contradictions between their findings, and produce a structured comparative summary. All six documents stay in context simultaneously, so Kimi draws connections across the full source set in a single pass rather than requiring the researcher to ask about each paper separately. This is a task where natural language processing capabilities combine directly with the context window advantage: understanding the semantic content of multiple long texts requires both.
Code Generation and Review
Consider a software engineer working with a large legacy codebase who needs to identify redundant functions, trace how a variable flows across 50 files, and suggest targeted refactors. The full codebase stays in context throughout the session, so Kimi does not lose sight of files uploaded at the beginning when the engineer asks about code added later. For document-heavy development tasks, Kimi's large context window can also reduce the need for a separate retrieval-augmented generation (RAG) pipeline in many use cases.
Writing Assistance and Editing
A technical writer submitting a full product manual for consistency review can work through the document section by section while Kimi retains the style guidelines and terminology established in the opening chapters. A novelist sharing a complete draft can ask Kimi to flag continuity issues in the final chapter against character details established in the first. Style context and narrative detail from early in the document remain accessible when Kimi is working on later sections.
Translation of Long-Form Documents
A legal translator working on a 150-page contract can rely on Kimi to maintain consistent terminology throughout the full document. The same defined term receives the same translation in clause 90 as it did in clause 2, rather than requiring the translator to re-establish definitions with each new section upload. This is particularly relevant for Chinese-to-English and English-to-Chinese translation, where Kimi's primary training language aligns with the source material.
Extended Q&A and Conversational Assistance
A product manager running a multi-session research sprint can establish project context, constraints, and prior findings at the start of a conversation, then continue refining strategy across 30 or more follow-up questions without re-explaining the project background each time. Earlier exchanges remain accessible throughout the session, so Kimi can reference context from the beginning of a long conversation without truncation.
How to Access Kimi: Pricing, Availability, and Language Support
Kimi is accessible without a paid subscription. The web interface at kimi.ai offers a free tier for general users, with separate paid options and API access available for higher-volume needs.
Is Kimi Free to Use?
Yes, Kimi offers a free tier at kimi.ai that requires no upfront payment. The free plan provides access to core AI assistant features including document upload and conversation. Usage limits may apply on the free tier; premium subscription options are available for users who need higher usage quotas or priority access. The API is priced separately on a per-token basis and is not part of the consumer free tier.
Is Kimi Available Outside China? Language Support
Kimi is accessible to international users via kimi.ai and supports both Chinese and English.
Users outside China can access Kimi through the web interface at kimi.ai and via the mobile app. Chinese is the primary language of Kimi's training data, but the product supports English-language conversations and document processing. Users outside China should verify current availability and any regional feature differences directly at kimi.ai, as access conditions may change over time.
How to Get Started with Kimi
Getting started with Kimi takes three steps:
- Visit kimi.ai in a web browser
- Create a free account using an email address or supported sign-in method
- Start a new conversation or upload a document to begin
Developers looking to integrate Kimi programmatically will find model specifications and API details in the next section.
Moonshot AI vs. ChatGPT and Claude: How Does Kimi Compare?
The table below compares Kimi, OpenAI's ChatGPT (GPT-4o), and Anthropic's Claude across the dimensions most relevant to a tool adoption decision: context window capacity, language support, availability, and pricing structure.
| Dimension | Kimi (Moonshot AI) | ChatGPT (GPT-4o) | Claude (Anthropic) |
|---|---|---|---|
| Context Window (Live Chat) | 128,000 tokens (~96,000 words) | 128,000 tokens (~96,000 words) | 200,000 tokens (~150,000 words) |
| Context Window (File Processing) | Up to 1,000,000 tokens (~750,000 words) | Up to 128,000 tokens (~96,000 words) | Up to 200,000 tokens (~150,000 words) |
| Primary Language Market | Chinese (English supported) | English (global) | English (global) |
| Global Availability | Accessible internationally via kimi.ai | Available globally | Available globally |
| Free Consumer Tier | Yes | Yes (GPT-4o, with limits) | Yes (with limits) |
| API Available | Yes (platform.moonshot.cn) | Yes (platform.openai.com) | Yes (api.anthropic.com) |
| Key Differentiator | Ultra-long file processing context (1M tokens) | Broad language support and third-party integrations | Large context with strong reasoning performance |
For users whose primary need is processing large documents in a single session, Kimi's 1 million token file processing capacity is the clearest factual advantage in this comparison. For users who need broad global language support, an extensive plugin and integration ecosystem, or the widest consumer adoption base, ChatGPT remains the stronger choice on those dimensions. Kimi K3: Moonshot AI's Flagship Reasoning Model provides useful context for evaluating where Kimi's latest models sit relative to the broader market.
Moonshot AI in the Chinese Competitive Landscape
Within China's domestic AI market, Moonshot AI competes against products backed by the country's largest technology companies. Baidu operates ERNIE Bot (文心一言), one of the first Chinese LLM products to reach consumers in 2023. Alibaba Cloud runs Qwen (通义千问), its own large language model and AI assistant. ByteDance (best known internationally for TikTok) offers Doubao (豆包) as its consumer AI product. Zhipu AI (智谱AI), an academic spinout with Tsinghua University roots, operates the ChatGLM model series, including notable open-source releases.
Against this field, Moonshot AI operates as an independent startup without the distribution advantages or platform scale of these incumbents. Its differentiated position rests on the long-context architecture rather than breadth of integration. Alibaba Group's investment in Moonshot AI adds an unusual dimension to this competitive picture: Alibaba is simultaneously a backer of Moonshot AI and the parent company of Alibaba Cloud, which operates the competing Qwen LLM.
For a head-to-head comparison of Moonshot AI vs. DeepSeek and Qwen, see Moonshot AI vs DeepSeek vs Qwen: China AI Models Compared.
Moonshot AI API: Developer Access and Model Specifications
Developers can integrate Kimi's language model capabilities directly into their applications through the publicly available Moonshot AI API, hosted at Moonshot AI's developer platform (platform.moonshot.cn). The API is OpenAI-compatible, meaning developers already familiar with OpenAI's chat completions endpoint can adopt the Moonshot AI API with minimal changes to their existing request structure. No need to relearn a fundamentally different interface.
The API offers three model variants, each suited to different context length requirements:
| Model Identifier | Context Window (Tokens) | Context Window (Words) | Recommended Use Case |
|---|---|---|---|
| moonshot-v1-8k | 8,000 tokens | ~6,000 words | Short conversations, quick Q&A, low-latency responses |
| moonshot-v1-32k | 32,000 tokens | ~24,000 words | Medium-length documents, extended multi-turn dialogue |
| moonshot-v1-128k | 128,000 tokens | ~96,000 words | Long documents, complex sessions, full-book processing |
The moonshot-v1-128k model covers the same 128,000-token context range as GPT-4o's API. For document-processing applications that need the full 1 million token capacity, verify current model availability and any specialized endpoints at the official Moonshot AI API documentation (platform.moonshot.cn/docs).
The API uses per-token pricing tiered by model. Pricing is subject to change; verify current rates at Moonshot AI's developer platform (platform.moonshot.cn) before integrating into a production application. To register for an API key, create a developer account at platform.moonshot.cn. Developers comparing the Moonshot AI API against OpenAI's will find the request structure familiar; the primary distinctions are the context window ceiling on the 128k model and the per-token pricing at each tier. For a comprehensive breakdown of API costs, see Moonshot Kimi API Pricing 2026: Plans and Cost Guide.
Moonshot AI Funding, Investors, and Valuation
Since its founding in March 2023, Moonshot AI has secured multiple rounds of venture capital funding from investors including Alibaba Group and HongShan (formerly Sequoia Capital China).
| Round | Date | Key Investors | Reported Valuation |
|---|---|---|---|
| Seed | 2023 | Undisclosed | Not publicly reported |
| Series A | 2023 | Multiple institutional investors | Not publicly reported |
| Series B | 2024 | Alibaba Group, HongShan | Approx. $2.5–$3B (reported per multiple technology news outlets, 2024) |
Valuation figures are based on publicly reported estimates and may not reflect the company's official disclosures. Verify specific figures against named sources and their publication dates before citing.
Alibaba Group's role in Moonshot AI's funding carries a dual significance: Alibaba is both an investor in Moonshot AI and the parent company of Alibaba Cloud, which operates the Qwen LLM, a direct competitor to Kimi in China's generative AI market. HongShan, which rebranded from Sequoia Capital China in 2023 after separating from the US entity, is one of China's most active technology venture investors.
For context on how Moonshot AI's reported valuation sits relative to its Western peers, Moonshot Kimi API Pricing 2026: Plans and Cost Guide illustrates the scale and commercial ambition of China's leading AI startups at their growth stage.
Moonshot AI in China's Generative AI Landscape
China's generative AI market is among the most active in the world, with major technology companies and funded startups releasing competing products in rapid succession since 2023. Established players include Baidu (ERNIE Bot), Alibaba Cloud (Qwen), ByteDance (Doubao), and Zhipu AI (ChatGLM), each backed by the distribution scale and capital resources of large technology groups. Moonshot AI occupies the position of an independent startup within this market, competing on product differentiation rather than platform reach. Its long-context specialization represents a specific bet on which capability gap matters most to professional and developer users.
Moonshot AI operates under China's Interim Measures for the Management of Generative Artificial Intelligence Services, issued by the Cyberspace Administration of China (CAC) and effective August 2023. These regulations require generative AI service providers serving users in China to register their services and comply with content standards, requirements that apply to all China-based AI providers, including Moonshot AI. The regulatory framework governs operational compliance in the domestic market and functions as the business operating context for China-based AI companies, comparable in structure to sector-specific AI regulations in other major markets.
Frequently Asked Questions About Moonshot AI
The questions below address the most frequently searched topics about Moonshot AI and Kimi, based on current search patterns.
What is Moonshot AI?
Moonshot AI (月之暗面, meaning "Dark Side of the Moon") is a Beijing-based Chinese generative AI startup founded in March 2023. The company builds Kimi, an AI assistant powered by a proprietary large language model with a context window of up to 1 million tokens for document processing. Moonshot AI is backed by investors including Alibaba Group and HongShan (formerly Sequoia Capital China).
Who founded Moonshot AI?
Moonshot AI was co-founded in March 2023 by Yang Zhilin (杨植麟), Zhang Yuhao, and Zhou Xinyu. Yang Zhilin serves as CEO and previously conducted AI research at Tsinghua University and Carnegie Mellon University, where he co-authored the XLNet language model paper (Yang et al., 2019), foundational research in modern language model pretraining approaches.
What is Moonshot AI's context window size?
Kimi's context window supports up to 128,000 tokens (approximately 96,000 words, roughly the length of a full novel) for live conversations. For file and document processing, the context window extends to 1 million tokens, equivalent to approximately 750,000 words. Verify current specifications at kimi.ai, as technical details may be updated.
Is Kimi available outside China, and does it support English?
Kimi is accessible to international users via kimi.ai and supports both Chinese and English. Chinese is the primary language of the model's training data, and English is confirmed as a supported language for both conversation and document processing. Users outside China should verify current regional availability directly at kimi.ai, as access conditions may change.
Is Kimi free to use?
Yes, Kimi offers a free tier at kimi.ai with no upfront cost. Usage limits may apply on the free plan, and premium options are available for higher-volume needs. API access via platform.moonshot.cn is billed separately on a per-token basis and is not included in the consumer free tier.
Who invested in Moonshot AI?
Confirmed investors in Moonshot AI include Alibaba Group and HongShan (formerly Sequoia Capital China). Moonshot AI was reported to be valued at approximately $2.5 billion to $3 billion as of 2024, according to multiple technology news outlets; verify specific figures against named sources and their publication dates. Alibaba Group also operates a competing LLM product (Qwen) through its Alibaba Cloud subsidiary.
How does Kimi compare to ChatGPT?
Kimi's primary factual advantage over OpenAI's ChatGPT (GPT-4o) is its document processing context window: up to 1 million tokens versus ChatGPT's ceiling of 128,000 tokens at the time of writing. ChatGPT offers broader global language support, a larger third-party integration ecosystem, and wider consumer adoption. See the comparison table in this article for a full side-by-side breakdown across context window size, availability, pricing model, and key differentiators.
Getting Started with Moonshot AI and Kimi
Moonshot AI builds Kimi around a single technical thesis: that the ability to process large volumes of text without truncation is a capability gap that matters to professional users. The product is accessible for free, supports English, and is open to international users.
To try Kimi as a consumer, visit kimi.ai and create a free account. The web interface allows you to start a conversation or upload documents immediately. No prior technical setup is required.
To access the Moonshot AI API as a developer, register at Moonshot AI's developer platform (platform.moonshot.cn) and review the official API documentation (platform.moonshot.cn/docs) for endpoint specifications, model identifiers, and current pricing.
Related Reading
- Kimi K3: Moonshot AI's Flagship Reasoning Model
- What Is Kimi K2? Moonshot AI's Open-Weight Thinking Model
- Moonshot Kimi API Pricing 2026: Plans and Cost Guide
- Moonshot AI vs DeepSeek vs Qwen: China AI Models Compared
- What Is Kimi AI? Guide to Moonshot AI
- MOONSHOTUSDT Perpetual Futures on Bybit
Last reviewed: 2025. Technical specifications, pricing, and availability details are subject to change. Verify current information at kimi.ai and platform.moonshot.cn before making adoption or integration decisions.