Term
Kimi K2
Kimi K2 is Moonshot AI's open-weight MoE model series with 1T total parameters and 32B active parameters, trained on 15.5T tokens — focused on agentic and coding tasks.
Kimi K2 — explained in more detail
Kimi K2 is the model family unveiled in summer 2025 by Chinese AI lab Moonshot AI. It is a Mixture-of-Experts model with 1 trillion total parameters and 32 billion active parameters per token, pretrained on 15.5 trillion tokens. Moonshot shipped two variants: Kimi-K2-Base for research and fine-tuning, and Kimi-K2-Instruct as a ready-to-use chat and agent model.
Where it fits
Kimi K2 is open-weight (available on Hugging Face) and forms the foundation for the follow-up generations K2.5 (February 2026, multimodal, 256k context) and K2.6 (April 2026). Within the Chinese open-source camp it sits alongside DeepSeek and Qwen as one of the central families for agentic coding workflows; western providers such as Cloudflare Workers AI and OpenRouter host the models as alternatives to GPT and Claude APIs.
Discover more
GPT-6 Astra Found Questions My First Security Review Missed
I used GPT-6 Astra and Fable 5.1 as independent reviewers for RLS, API and tenant-isolation checks. The useful part was the structured cross-review.
GlossaryKimi K3
Kimi K3 (July 2026) is Moonshot AIs open-weight flagship — an MoE model with 2.8T parameters (104B active), native vision and a 1M-token context.
EncyclopediaKimi (Moonshot AI) — the open model line at a glance
Kimi by Moonshot AI as an open line: positioning, use profile and how to choose. Flagship Kimi K3 with a very large context window.