FitMyLLM
Moonshot AI/Mixture of Experts

Moonshot AIKimi K2.7 Code

Kimi K2.7 Code — the coding-specialised K2.7: 1T total, 32B active, MLA attention, 256K context.

chatcodingreasoningagentictool_usemultilingual
1026.88B
Parameters (32B active)
256K
Context length
11
Benchmarks
17
Quantizations
568K
HF downloads
Architecture
MoE
Released
2026-06-12
Layers
61
KV Heads
64
Head Dim
112
Family
kimi

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
IQ2_XXS2.38
306.8 GB
306.0 + 0.8 KV
low
IQ2_M2.93
377.4 GB
376.6 + 0.8 KV
low
Q2_K3.16
406.9 GB
406.1 + 0.8 KV
low
IQ3_XXS3.25
418.5 GB
417.7 + 0.8 KV
low
IQ3_XS3.5
450.6 GB
449.7 + 0.8 KV
low
Q3_K_S3.64
468.5 GB
467.7 + 0.8 KV
low
IQ3_M3.76
483.9 GB
483.1 + 0.8 KV
low
Q3_K_M4
514.7 GB
513.9 + 0.8 KV
low
Q3_K_L4.3
553.2 GB
552.4 + 0.8 KV
moderate
IQ4_XS4.46
573.8 GB
573.0 + 0.8 KV
moderate
Q4_K_S4.67
600.7 GB
599.9 + 0.8 KV
moderate
Q4_K_M4.89
629.0 GB
628.2 + 0.8 KV
good
Q5_K_S5.57
716.3 GB
715.5 + 0.8 KV
good
Q5_K_M5.7
732.9 GB
732.1 + 0.8 KV
good
Q6_K6.56
843.3 GB
842.5 + 0.8 KV
excellent
Q8_08.5
1092.4 GB
1091.5 + 0.8 KV
lossless
FP1616
2055.1 GB
2054.2 + 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 ~307 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 K2.7 Code now

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

Community Ratings

Loading ratings...

Benchmarks (11)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run hf.co/unsloth/Kimi-K2.7-Code-GGUF:Q4_K_M

Not in Ollama's library: Ollama pulls the GGUF from Hugging Face (unsloth/Kimi-K2.7-Code-GGUF). If Q4_K_M is not in that repo, pick a file listed there.

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

Find the best GPU for Kimi K2.7 Code

Build Hardware for Kimi K2.7 Code
Spec sheet

Kimi K2.7 Code — 1026.88B MoE.

Specifications
Parameters
1026.88B (32B active)
Architecture
Mixture of Experts
Context Length
256K tokens
Capabilities
chat, coding, reasoning, agentic, tool_use, multilingual
Release Date
2026-06-12
Provider
Moonshot AI
Family
kimi
VRAM requirements
QuantBPWVRAMQuality
IQ2_XXS2.38306.0 GB65%
IQ2_M2.93376.6 GB75%
Q2_K3.16406.1 GB78%
IQ3_XXS3.25417.7 GB82%
IQ3_XS3.5449.7 GB84%
Q3_K_S3.64467.7 GB85%
IQ3_M3.76483.1 GB86%
Q3_K_M4513.9 GB88%
Q3_K_L4.3552.4 GB90%
IQ4_XS4.46573.0 GB92%
Q4_K_S4.67599.9 GB93%
Q4_K_M4.89628.2 GB94%
Q5_K_S5.57715.5 GB96%
Q5_K_M5.7732.1 GB96%
Q6_K6.56842.5 GB97%
Q8_08.51091.5 GB100%
FP16162054.2 GB100%
Benchmark scores
GPQA Diamond89.6
HLE35.0
AA Intelligence43.0
AA Coding60.8
IFBench63.1
Terminal-Bench Hard44.7
τ²-Bench90.1
SciCode47.5
AA Long Context75.0
Terminal-Bench 2.167.4
Terminal-Bench 41.0