Moonshot AI (Kimi)
Open-weights agentic frontier models trained cheaply, with Kimi Delta Attention (linear attention), the Muon optimizer, Agent Swarm parallel sub-agents, and 1M context at 2.8T parameters (K3). The stated aim is to convert energy into intelligence optimally.
- Founded
- March 2023
- Headquarters
- Beijing, China
- Founders
- Yang Zhilin, Zhou Xinyu, Wu Yuxin
- Pushes against
- Closed US frontier labs; full-attention transformers.
- Now
- ARR passed $200M in April and reportedly $300M after K3. Reported (Reuters, 2026-09-03) to have filed confidentially for a Hong Kong IPO; pre-IPO round targeting up to $50B reported (Bloomberg).
Key people
- Yang Zhilin (founder, CEO)
Funding
- July 2026
- Amount
- undisclosed (closed above goal)
- Valuation
- $31.5B-$35B (reports differ)
- May 2026
- Amount
- $2B
- Valuation
- $20B
- Lead
- Long-Z Investments (Meituan), with Tsinghua Capital, China Mobile
Every funding round in the atlas
Latest launches
- Kimi K3 Moonshot AI · 16 July 2026
2.8T-parameter open-weights MoE (104B active, 1M context) on Kimi Delta Attention and Attention Residuals, reported third on Artificial Analysis behind two closed models at launch. - Kimi K2.7 Code Moonshot AI · 12 June 2026
Coding-only K2.6 derivative with better instruction-following in long contexts and about 30% less overthinking; thinking mode always on. - Kimi K2.6 Moonshot AI · 20 April 2026
Coding-focused K2 update claiming 80.2% SWE-bench Verified and agent swarms of 300 sub-agents over 4,000 coordinated steps, up from 100 and 1,500. - Attention Residuals (AttnRes) Moonshot AI (Kimi Team) · 16 March 2026
Replaces fixed residual accumulation with softmax attention over earlier layers' outputs, fixing PreNorm dilution; Block AttnRes keeps memory manageable. - Kimi K2.5 Moonshot AI · 27 January 2026
Natively multimodal 1T/32B open model trained on about 15T mixed vision-text tokens, with Agent Swarm orchestrating up to 100 parallel sub-agents. - Kimi K2 Thinking Moonshot AI · 6 November 2025
Open-weights thinking agent that interleaves reasoning with 200-300 sequential tool calls; claims state of the art on HLE with tools (44.9%) and BrowseComp (60.2%). - Kimi Linear Moonshot AI · 30 October 2025
Hybrid linear-attention architecture (Kimi Delta Attention + MLA) that beats full attention in fair comparisons while cutting KV cache up to 75%. - Kimi K2-Instruct-0905 Moonshot AI · 5 September 2025
K2 refresh that doubles context from 128K to 256K and improves agentic coding, with SWE-bench Verified rising from 65.8 to 69.2 and Terminal-Bench from 37.5 to 44.5.
Launches by year
19 launches and papers since 9 October 2023, oldest first within each year.
- 2026 5
- Kimi K2.5, Attention Residuals (AttnRes), Kimi K2.6, Kimi K2.7 Code, Kimi K3
- 2025 11
- Kimi k1.5, Muon is Scalable for LLM Training (Moonlight), Moonlight / Muon is Scalable, Kimi-VL, Kimi-Dev-72B, Kimi-Researcher, Kimi K2, Kimi K2 technical report (MuonClip), Kimi K2-Instruct-0905, Kimi Linear, Kimi K2 Thinking
- 2024 2
- Kimi 2M-character context beta, k0-math
- 2023 1
- Kimi Chat
Sources
- kimi.com/blog
- kimi.com/blog/kimi-k3
- techcrunch.com/2026/05/07/chinas-moonshot-ai-raises-2b-at-20b-valuation-as-demand-for-open
- startupfortune.com/moonshot-ais-kimi-k3-model-sends-its-valuation-toward-50-billion-and-a-
- en.wikipedia.org/wiki/Moonshot_AI
- dealstreetasia.com/stories/moonshot-hong-kong-ipo-494042/
This profile was checked against its sources on 6 October 2026. How we check