ai-toolkit/model-producers/moonshot-ai

Kimi AI Research Labs

Value Proposition & Features

Moonshot AI (北京月之暗面科技有限公司) is a Beijing-based artificial intelligence company that develops large language models and multimodal AI, best known for its Kimi chatbot and Kimi K2 open-weight model family. [du1dmk] [m5p6tw] [v02z8j] It positions itself as a leading generative AI and LLM provider focused on long‑context reasoning, agentic workflows, and consumer “super app” experiences for general users, developers, and enterprises. [du1dmk] [m5p6tw] [v02z8j]
Core product features (Kimi / Moonshot platform):
  • Consumer AI assistant (Kimi Chat): Web and mobile chatbot at kimi.com that supports long-form conversations, online search, multimodal reasoning (text + images), deep thinking, and complex task workflows in English and Chinese. [92di2m] [m5p6tw] [v02z8j]
  • Open‑weight model family (Kimi K2.x): Trillion‑parameter LLMs such as Kimi K2.6 and K2.7 Code, released as open weights for self‑hosting while also powering Moonshot’s hosted chat, API, and tools. [92di2m] [vl431q] [zayn9u] [m5p6tw]
  • Developer & agent platform: OpenAI‑compatible API at api.moonshot.ai, plus tools like Kimi Work desktop agents and Kimi Code CLI that enable agent swarms, coding automation, and integration into applications and data workflows. [92di2m] [vl431q] [asr2p9] [vt6z8f] [v02z8j]
Key features (priority order):
  • Kimi Chat assistant – consumer web and mobile AI assistant for chat, document analysis, coding, and web search via kimi.com and Kimi apps. [92di2m] [na6pt5] [m5p6tw] [v02z8j]
  • Long‑context LLMs – Kimi models specialized for handling extensive text (legal, finance, creative writing) with long context windows and deep reasoning. [du1dmk] [sdwyh2] [m5p6tw] [v02z8j]
  • Open‑weight Kimi K2.x models – flagship Kimi K2.6 and K2.7 Code released as open weights on platforms like Hugging Face and GitHub for self‑hosting and customization. [92di2m] [vl431q] [zayn9u] [m5p6tw]
  • OpenAI‑compatible API – international Kimi API exposing a Chat Completions interface at https://api.moonshot.ai/v1/chat/completions for developers. [vl431q] [asr2p9] [vt6z8f]
  • Agentic workflows & Agent Swarm – Kimi Work and Kimi agent modes support up to ~300 sub‑agents with browser automation for multi‑step task execution. [92di2m] [vl431q]
  • Desktop & CLI tools – Kimi Work desktop agent for macOS/Windows and Kimi Code CLI for code‑centric workflows and local task automation. [92di2m] [vl431q] [v02z8j]
  • Multimodal AI – models and tools with vision capabilities and multimodal reasoning (text + images). [m5p6tw] [v02z8j]
  • Enterprise & research tooling – tools like Kimi Researcher and integrations for data workflows, optimized for professional and enterprise use. [v02z8j]

Screenshots

No reliable source found for official Moonshot AI or Kimi product screenshots that can be directly hot‑linked; primary sites either require login or use dynamic assets without stable public image URLs. [92di2m] [na6pt5] [m5p6tw] [v02z8j]

Product Roadmap / Announcements

As of July 20, 2026,
  • 2026‑06 – Moonshot AI launched Kimi Work desktop agent for macOS and Windows testing, combining a 300‑subagent Agent Swarm with browser automation to run tasks directly on users’ machines. [92di2m]
  • 2026‑05 – Kimi K2.6 became the flagship 1‑trillion‑parameter open‑weight model with Agent Swarm capabilities, available via free chat at kimi.com, paid subscriptions, developer API, and open weights distribution. [92di2m] [vl431q]
  • 2026‑05 – International Kimi API documented with OpenAI‑compatible Chat Completions interface at https://api.moonshot.ai/v1, indicating a stable developer platform and ongoing API evolution. [vt6z8f]
  • 2026‑04–05 – Launch and promotion of Kimi Code CLI and expanded open‑weight releases (K2.5 updated with branches for K2.6), positioning Moonshot as a major open‑source model provider. [vl431q] [m5p6tw]

Recent Developments

  • In mid‑2026, Kimi K2.6 was highlighted by multiple technical blogs as an open‑source AI agent that rivals or “beats” leading frontier models like GPT‑5.4 in some benchmarks, emphasizing Moonshot’s role in the open‑weight LLM ecosystem. [vl431q] [m5p6tw]
  • Analyst and developer coverage in 2025–2026 described Moonshot AI as “one of the loudest names in open AI,” with Kimi models topping Hugging Face download charts and gaining traction among international developers. [92di2m] [m5p6tw]
  • By May 2026, reports cited Moonshot AI’s valuation at around $18–20B, reflecting rapid growth and investor confidence in its Kimi platform and open‑weight strategy. [d15elw] [m5p6tw]

History and Origin Story

Moonshot AI (Beijing Moonshot AI Technology Co., Ltd.) is a private AI company incorporated in Haidian District, Beijing, co‑founded in March 2023 by Tsinghua University alumni Yang Zhilin, Zhou Xinyu, Wu Yuxin, and (per some sources) Zhang Yutao. [du1dmk] [m5p6tw] [v02z8j] The company was named “Moonshot AI” after Pink Floyd’s The Dark Side of the Moon, reflecting founder Yang Zhilin’s vision of exploring the “dark side of the moon” of AI and converting energy into intelligence, and it quickly became known for long‑context LLMs and the Kimi chatbot. [du1dmk] [m5p6tw] [v02z8j]

Fundraising History

Conflicting funding data exists; more recent aggregated sources converge on total funding around the high hundreds of millions of USD, while some databases list multi‑billion figures likely including later rounds or broader capital structures. [d15elw] [sdwyh2] [m5p6tw] [v02z8j]
RoundDateAmountLead investor
Seed / Early2023Not disclosedAlibaba (largest outside investor reported) [m5p6tw] [v02z8j]
Series B–D2024–2026Not fully disclosedMeituan, Zhen Fund, Capital Today (reported major investors) [sdwyh2] [v02z8j]
Latest2025–2026Part of total ≈$510M–$3.77B (figures vary by source) [d15elw] [sdwyh2] [m5p6tw] Undisclosed / multiple
RoundDateAmountLead investor
Total2023–2026≈$510M–$3.77B reported across sources [d15elw] [sdwyh2] [m5p6tw] [v02z8j] Multiple
Investors (alphabetical):
  • Alibaba [m5p6tw] [v02z8j]
  • Capital Today [sdwyh2]
  • Meituan [sdwyh2]
  • Zhen Fund [sdwyh2]

Notable Team Members

Yang Zhilin (Founder & CEO) – Tsinghua University alumnus and CEO who named Moonshot AI after The Dark Side of the Moon; he leads the company’s strategy around long‑context LLMs and consumer super apps powered by Kimi. [m5p6tw] [v02z8j]
Zhou Xinyu (Co‑founder) – Tsinghua alumnus and co‑founder associated with Moonshot’s platform and LLM development; cited among the founding team in multiple company profiles. [du1dmk] [sdwyh2] [m5p6tw] [v02z8j]
Wu Yuxin (Co‑founder) – Also from Tsinghua University, recognized as a co‑founder involved in Moonshot AI’s early technical and product direction for Kimi and long‑context models. [du1dmk] [sdwyh2] [m5p6tw] [v02z8j]
Zhang Yutao (Co‑founder, per some sources) – Listed by some databases and glossaries as a co‑founder of Moonshot AI, contributing to the initial formation of Beijing Moonshot AI Technology Co., Ltd. [du1dmk] [sdwyh2]

Market Sizing

Category, Market Size, and Category Growth

Moonshot AI operates in the generative AI / large language model category, specifically as a model vendor and producer of foundation LLMs and multimodal models, and as a consumer AI assistant platform via Kimi. [du1dmk] [m5p6tw] [v02z8j] It participates in the global generative AI market that major analyst firms project to reach hundreds of billions of dollars in the 2030s, driven by rapid adoption of LLMs in consumer, enterprise, and developer tooling; Moonshot’s focus on long‑context and open‑weight models positions it within the high‑growth subsegment of open and agentic LLM platforms. [du1dmk] [sdwyh2] [m5p6tw] [v02z8j]

Pricing

Available API pricing information for Kimi models via Moonshot’s platform (international developer usage):
Tier / ModelPrice (input)Price (output)Notes
Kimi K2.5 (API)$0.60 / 1M tokens$3.00 / 1M tokensListed as the cheapest current model for API usage. [asr2p9]
Other Kimi modelsNot publicly standardizedNot publicly standardizedDetailed public pricing beyond K2.5 not reliably documented. [asr2p9] [zayn9u] [vt6z8f]
If accessing Kimi via consumer chat (web/app), sources mention a free tier and premium subscriptions but do not provide explicit public price points. [92di2m] [m5p6tw] Overall, there is partial public pricing (API K2.5) and otherwise “no public pricing” for full tier details.

Revenue Trajectory Estimates

No reliable source found with specific revenue or ARR figures for Moonshot AI; available data focuses on valuation and funding rather than revenue metrics. [d15elw] [sdwyh2] [m5p6tw] [v02z8j]

Competitive Landscape

Who it's for, who it's not for

Moonshot AI and Kimi are designed for general users, developers, and enterprises seeking an AI assistant and LLM platform that excels at very long‑context tasks (legal documents, finance analysis, creative writing), multimodal reasoning, and agentic workflows, with both hosted chat and open‑weight deployment options. [du1dmk] [92di2m] [sdwyh2] [m5p6tw] [v02z8j] It particularly suits users who need Chinese and English support, want to self‑host powerful open‑weight models, or integrate an OpenAI‑compatible API and agent tools like Kimi Work and Kimi Code into applications and data workflows. [92di2m] [vl431q] [asr2p9] [vt6z8f] [m5p6tw] [v02z8j]
Moonshot AI may be less suitable for organizations that require exclusively closed proprietary frontier models from US Hyperscalers, or that need deeply integrated services in ecosystems tightly bound to other cloud vendors. [92di2m] [zayn9u] [m5p6tw]
It is also less aligned with users who prefer simple, low‑control chatbot experiences without interest in long‑context, agent swarms, or developer APIs, since much of its differentiation is in advanced LLM features and tooling. [92di2m] [vl431q] [m5p6tw] [v02z8j]

Viable Alternatives

  • OpenAI (ChatGPT / GPT‑4/5 family) – Competing closed‑weight LLMs and chat assistants with strong global ecosystem and tooling, an obvious alternative for general AI assistant and API use. [92di2m] [vl431q] [zayn9u] [m5p6tw]
  • Anthropic (Claude) – Focus on safety and reasoning; Claude Opus is frequently used as a benchmark for agentic coding quality against Kimi K2.7 Code. [zayn9u]
  • Google (Gemini) – Multimodal LLMs integrated into Google Workspace and cloud, competing in both consumer and enterprise AI assistant categories. [zayn9u] [m5p6tw]
  • Meta (LLaMA open‑weight models) – Open‑weight LLM family widely used for self‑hosting and customization, similar to Kimi K2 open‑weight strategy. [92di2m] [zayn9u] [m5p6tw]
  • Alibaba / Tongyi Qianwen – Major Chinese LLM provider and reported largest external investor in Moonshot, offering alternative models and enterprise AI services in the same regional market. [m5p6tw]

Competitor Table

CompetitorDescription
OpenAIProvider of ChatGPT and GPT‑4/5‑series closed‑weight models with a dominant global API and consumer assistant platform competing directly with Kimi for chat and developer use. [92di2m] [vl431q] [zayn9u] [m5p6tw]
AnthropicCreator of Claude models (e.g., Claude Opus), often used as a reference point when evaluating Kimi K2.7 Code’s agentic coding quality and safety‑focused reasoning capabilities. [zayn9u]
[Google Gemini](Google Gemini)Google’s multimodal LLM suite integrated across consumer products and Google Cloud, offering chat, coding, and enterprise AI similar to Moonshot’s Kimi ecosystem. [zayn9u] [m5p6tw]
[Meta Llama](Meta Llama)Family of open‑weight LLMs that enable self‑hosting and customization, representing a major alternative for developers favoring open models like Kimi K2.x. [92di2m] [zayn9u] [m5p6tw]
[Alibaba / Tongyi Qianwen](Alibaba Tongyi Qianwen)Alibaba’s LLM platform and enterprise AI offering in China; as Moonshot AI’s largest outside investor, it still competes in providing generative AI services to regional enterprises. [m5p6tw]

Sources