Moonshot AI/Mixture of Experts

MKimi K2.6

Name: Kimi K2.6
Author: Moonshot AI

Kimi K2.6 is an open-source, native multimodal agentic model that advances practical capabilities in long-horizon coding, coding-driven design, proactive autonomous execution, and swarm-based task orchestration.

chatcodingreasoningmultilingualvisionmathagentictool_use

1058.59B

Parameters (32B active)

256K

Context length

Benchmarks

Quantizations

Architecture

MoE

Released

2026-04-20

Layers

KV Heads

Head Dim

112

Family

kimi

Quantization Options

Context length:

Quant	Bits	VRAM @ 16K	Quality
IQ2_XXS	2.38	316.2 GB 315.4 + 0.8 KV	low
IQ2_M	2.93	389.0 GB 388.2 + 0.8 KV	low
Q2_K	3.16	419.4 GB 418.6 + 0.8 KV	low
IQ3_XXS	3.25	431.3 GB 430.5 + 0.8 KV	low
IQ3_XS	3.5	464.4 GB 463.6 + 0.8 KV	low
Q3_K_S	3.64	483.0 GB 482.1 + 0.8 KV	low
IQ3_M	3.76	498.8 GB 498.0 + 0.8 KV	low
Q3_K_M	4	530.6 GB 529.8 + 0.8 KV	low
Q3_K_L	4.3	570.3 GB 569.5 + 0.8 KV	moderate
IQ4_XS	4.46	591.5 GB 590.7 + 0.8 KV	moderate
Q4_K_S	4.67	619.2 GB 618.4 + 0.8 KV	moderate
Q4_K_M	4.89	648.4 GB 647.6 + 0.8 KV	good
Q5_K_S	5.57	738.3 GB 737.5 + 0.8 KV	good
Q5_K_M	5.7	755.5 GB 754.7 + 0.8 KV	good
Q6_K	6.56	869.3 GB 868.5 + 0.8 KV	excellent
Q8_0	8.5	1126.0 GB 1125.2 + 0.8 KV	lossless
FP16	16	2118.5 GB 2117.7 + 0.8 KV	lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Too big for a single GPU — plan a multi-GPU deployment

Even the lightest quant needs ~316 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →

▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN KIMI K2.6 NOW

Rent on Vast.ai →Or RunPod →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (9)

τ²-Bench95.9

GPQA Diamond91.1

IFBench76.0

AA Long Context69.7

AA Intelligence53.9

SciCode53.5

AA Coding47.1

Terminal-Bench43.9

HLE35.9

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run kimi:1059b-q4_K_M

Tag may need adjustment — check ollama.com/library/kimi for available tags.

▸ SETUP GUIDE

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more

Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

HuggingFace GGUF Downloads Build Hardware

Find the best GPU for Kimi K2.6

Build Hardware for Kimi K2.6

QUANT	BPW	VRAM	QUALITY
IQ2_XXS	2.38	315.4 GB	65%
IQ2_M	2.93	388.2 GB	75%
Q2_K	3.16	418.6 GB	78%
IQ3_XXS	3.25	430.5 GB	82%
IQ3_XS	3.5	463.6 GB	84%
Q3_K_S	3.64	482.1 GB	85%
IQ3_M	3.76	498.0 GB	86%
Q3_K_M	4	529.8 GB	88%
Q3_K_L	4.3	569.5 GB	90%
IQ4_XS	4.46	590.7 GB	92%
Q4_K_S	4.67	618.4 GB	93%
Q4_K_M	4.89	647.6 GB	94%
Q5_K_S	5.57	737.5 GB	96%
Q5_K_M	5.7	754.7 GB	96%
Q6_K	6.56	868.5 GB	97%
Q8_0	8.5	1125.2 GB	100%
FP16	16	2117.7 GB	100%