# Moonshot AI Moonshot AI (月之暗面; "Moonshot" / "Dark Side of the Moon") is a Chinese [[AI Frontier Model|frontier AI]] lab based in Beijing, founded in 2023 by [[Yang Zhilin]] (杨植麟), a former researcher at Meta AI and Google Brain. The company is best known for the [[Kimi]] family of [[Large Language Models (LLMs)]] and Kimi-branded consumer and developer products. Moonshot is one of the four "AI Tigers" of China alongside [[Deepseek]], [[Zhipu AI (Z.ai)]], and [[Baichuan]], and has raised funding from Alibaba, Tencent, Sequoia China, and others. It ships both hosted products and [[AI Open Weight Models|open-weight]] releases, positioning itself as a cost-performance alternative to US closed-weight labs ([[OpenAI]], [[Anthropic]], [[Google DeepMind]]) on agentic and long-context workloads. ## Products - **Kimi** — consumer chatbot and general-purpose AI assistant (kimi.com + Kimi App) - **Kimi K2 family** — open-weight [[AI Mixture of Experts (MoE)|MoE]] models; see [[Kimi K2.6]] (latest) and [[Kimi K2.5]] (prior generation, referenced in [[AI Expert Offloading]]) - **[[Kimi Code]]** — coding-specific product / CLI surface built on top of Kimi K2 - **Kimi API** (platform.kimi.ai) — developer access - **Claw Groups** (research preview) — multi-agent, multi-human collaboration platform announced alongside K2.6 ## Positioning - Long-context and agentic coding as the primary differentiation axis, not chat quality - Open-weight releases for the K2 model line; closed for some serving infrastructure - Heavy investment in long-horizon autonomy: 4,000+ tool calls over 12+ hours, 300-agent swarms (K2.6) - Competes with [[Claude Opus 4.7]], [[GPT-5.4]], and [[Gemini 3]] on agentic coding benchmarks (Terminal-Bench 2.0, SWE-Bench Pro, BrowseComp) ## Why it matters - Signals that Chinese labs can ship frontier-competitive *agentic* models, not just cheaper chat models - Open-weight Kimi K2 weights undercut the "you need our hosted API for serious agent work" moat held by US labs - Part of a broader 2026 shift where the frontier is measured in duration of autonomy and tool-call stability rather than raw reasoning benchmarks ## References - Official website: https://www.kimi.com/ - Blog: https://www.kimi.com/blog - Kimi API platform: https://platform.kimi.ai - Wikipedia: https://en.wikipedia.org/wiki/Moonshot_AI ## Related - [[Yang Zhilin]] - [[Kimi]] - [[Kimi K2.6]] - [[Kimi K2.5]] - [[Kimi Code]] - [[Kimi CLI]] - [[Large Language Models (LLMs)]] - [[AI Frontier Model]] - [[AI Open Weight Models]] - [[AI Mixture of Experts (MoE)]] - [[Deepseek]] - [[Zhipu AI (Z.ai)]] - [[Qwen]] - [[OpenAI]] - [[Anthropic]] - [[Google DeepMind]]