Kimi
AI assistant from Moonshot AI for web search and deep research, and for creating documents, slides, spreadsheets and websites, on web and mobile.
Official Website ↗
Details and Sources
Status: activeReviewed: 2026-10-14
Moonshot AI Funding History
Moonshot AI is a Chinese AI company developing the Kimi large language models and assistant. Its work combines foundation-model development with an assistant for search, research and content creation.
Company Overview
Founded 2023 · Beijing, China
Official WebsiteMoonshot AI Funding Timeline
Series F
Over $3.5 billion · $35B valuation
Led by National AI Industry Investment Fund.
Source · en.tmtpost.com ↗Meituan-Led Round (Unlabeled; Media Call It Series D/E)
~US$2 billion
Led by Long-Z Investments.
Source · bloomberg.com ↗Series C (Media Label)
US$500 million
Led by IDG Capital. Participation included Alibaba Group, Tencent.
Source · scmp.com ↗
Moonshot AI Founding Team
Yang Zhilin, Zhou Xinyu and Wu Yuxin
Moonshot AI Products
Applications, APIs and Tools
Kimi Claw
Kimi's one-click cloud hosting of the OpenClaw personal AI assistant, bundling models and skills for always-on operation.
Official Website ↗Details and Sources
Status: activeReviewed: 2026-10-14
Models
Kimi
Moonshot AI's language model family for text, code and reasoning tasks, including vision-capable and open-weight releases from K1.5 through K3.
Official Website ↗
Details and Sources
Status: activeReviewed: 2026-10-18
Recorded Versions · 6
Kimi K1.5
Multimodal reasoning model from Moonshot AI, released with a technical report.
Official Website ↗
Details and Sources
Status: activeReviewed: 2026-10-14
Kimi K2
Open-weights mixture-of-experts model released under a modified MIT license.
- Parameters
- 1T total / 32B active (MoE)(2026-10-14)
Details and Sources
Status: activeReviewed: 2026-10-14
Sources:en.wikipedia.orghuggingface.co
Kimi K2.5
Open-weights multimodal model processing text, images and video with agent-swarm execution.
Official Website ↗
Details and Sources
Status: activeReviewed: 2026-10-14
Kimi K2.6
Open-weight mixture-of-experts language model from Moonshot AI; a native multimodal agentic release exposing Thinking and Instant inference modes through a single checkpoint.
- Parameters
- 1T total / 32B active parameters (MoE)(2026-10-14)
- Context Window
- 256K tokens(2026-10-14)
- License
- Modified MIT(2026-10-14)
Details and Sources
Status: activeReviewed: 2026-10-17
- Modalities
- Text, image and video input; text output(2026-10-14)
- API Access
- API model id kimi-k2.6 on the Kimi platform; kimi-k2.5 listed as deprecated(2026-10-14)
Kimi K2 Thinking
Reasoning-focused open-weights update to Kimi K2, supporting interleaved chain-of-thought with tool calls.
Official Website ↗
Details and Sources
Status: activeReviewed: 2026-10-14
Kimi K3
Open-weights mixture-of-experts model whose weights were published following the release announcement.
- Parameters
- 2.8T total (MoE)(2026-10-14)
Details and Sources
Status: activeReviewed: 2026-10-14
Sources:en.wikipedia.orghuggingface.co
FAQ
How should Moonshot AI and Kimi be distinguished?
Moonshot AI is the company identity used for the financing record. Kimi is its model and assistant brand; product usage and company financing are separate measures.
Does a longer context window guarantee better research answers?
No. Retrieval quality, attention to relevant details and source verification still matter. Capacity to accept more text is not a guarantee of correct reasoning.
Why should model benchmarks be compared by release version?
Capabilities and evaluation settings can change between releases. A score without the model version and test conditions is difficult to interpret.
Does Moonshot AI’s Series F valuation establish profitability?
No. A financing valuation is a transaction price for equity. It does not establish operating profit or sustainable cash generation.
How does an assistant subscription differ from API consumption?
A subscription packages access for end users, while API consumption supports applications through programmatic requests. Their prices and usage limits need separate interpretation.
What matters when evaluating an assistant’s search citations?
Check whether each cited page supports the associated claim, whether it is current enough and whether a primary source is available.