Term
Kimi K2.6
Kimi K2.6 is Moonshot AI's open-weight agent model from April 2026 — 1T parameters (32B active), 262k context, natively multimodal, with agent swarms of up to 300 sub-agents.
Kimi K2.6 — explained in more detail
Moonshot AI shipped Kimi K2.6 on April 20, 2026 as a general-availability model built for 12-hour coding sessions, agent swarms and full-stack generation. The MoE model keeps the 1T / 32B-active architecture of the K2 family, extends the context window to 262,144 tokens and delivers native multimodality for text, images and video — without separate vision modules. It ships natively in INT4 quantization.
Example / Practical use
On agentic benchmarks, K2.6 posts competitive numbers against GPT-5.4 and Claude Opus 4.6: BrowseComp 83.2, SWE-Bench Verified 80.2, Terminal-Bench 2.0 66.7. For complex tasks the model can launch up to 300 sub-agents in parallel and coordinate them across up to 4,000 steps. Available through Kimi.com, the Kimi App, the official API, the Kimi Code CLI as well as third parties like Cloudflare Workers AI and OpenRouter.
Entdecke mehr
Three Days with Claude Fable 5 — and Why Opus Stays My Daily Driver
Fable 5 fixed a bug in one shot that Opus failed at twice. I'm still not switching. My honest impressions from the trenches after three days.
GlossarKimi 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.
LexikonComparing AI models — who builds what and how to choose
The major model families in 2026 at a glance. Who builds Claude, GPT, Gemini, Llama, Mistral, DeepSeek, Qwen — and which model to pick when.