AtomicChat / Kimi-K3-GGUF

huggingface.co
Total runs: 6.9K
24-hour runs: -54
7-day runs: 356
30-day runs: -1.5K
Model's Last Updated: July 31 2026
image-text-to-text

Introduction of Kimi-K3-GGUF

Model Details of Kimi-K3-GGUF


Kimi K3

Kimi K3 , self-quantized to GGUF by Atomic Chat . Built straight from Moonshot AI's original weights with a per-tensor importance matrix, so this is not a repack of somebody else's files. Runs fully offline.

Highlights
  • 2779.9B parameters : the weights this repo quantizes.
  • Context length : 1,048,576 tokens (1M), as published by Moonshot AI.
  • 93 layers : Mixture-of-Experts.
  • Modalities : the base model handles Text, Image; this repo ships text-only quants, it carries no vision projector.
  • Full imatrix ladder : every quant is calibrated with an importance matrix.
  • New Architecture : Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and scales up MoE sparsity with a Stable LatentMoE framework that activates 16 out of 896 experts, yielding an approximate 2.5× improvement in overall scaling efficiency over Kimi K2.
  • Long-Horizon Coding : Operating with minimal human oversight, Kimi K3 sustains long engineering sessions, navigates massive repositories, and orchestrates terminal tools, from GPU kernel optimization and compiler development to vision-in-the-loop game dev, CAD, and even chip design.

These GGUFs are self-quantized from the original weights , not a repack. The importance matrix keeps low-bit quants closer to the full-precision model.

Always pass --jinja so the Kimi K3 chat template is applied. Without it the model can emit malformed turns.

Model Overview
Property Value
Base model moonshotai/Kimi-K3
Parameters 2779.9B
Layers 93
Experts 896 routed
Context length 1,048,576 tokens (1M)
Vocabulary 163,840
Modalities Text, Image in the base model; text only in this repo, it ships no vision projector
Architecture Mixture-of-Experts, 896 experts, 96 attention heads over 96 KV heads, KimiK3ForConditionalGeneration
This repo GGUF quants (imatrix)
Get started

Run Kimi K3 locally with:

  • Atomic Chat : the easiest path. Open the app, search AtomicChat/Kimi-K3-GGUF , pick a quant, hit Use this model .
  • llama.cpp: llama-server -hf AtomicChat/Kimi-K3-GGUF:None --jinja -c 8192
  • Ollama: ollama run hf.co/AtomicChat/Kimi-K3-GGUF:None
  • LM Studio / Jan: search the repo id, download any quant.
Best practices
Parameter Value
temperature 1.0.

Moonshot AI's recommended sampling configuration for moonshotai/Kimi-K3 .

Run in llama.cpp
git clone https://github.com/ggml-org/llama.cpp
cmake llama.cpp -B llama.cpp/build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
cmake --build llama.cpp/build --config Release -j --target llama-cli llama-server
./llama.cpp/build/bin/llama-server \
    -hf AtomicChat/Kimi-K3-GGUF:None \
    --jinja -ngl 99 -c 8192 -fa on
How these were made
  1. Download moonshotai/Kimi-K3 (original weights).
  2. Convert to f16 GGUF with llama.cpp .
  3. Build an importance matrix over our calibration corpus.
  4. Quantize the ladder with --imatrix .
License

Original model by Moonshot AI, released under the other license. Full terms: other . Quantized by Atomic Chat.

Runs of AtomicChat Kimi-K3-GGUF on huggingface.co

6.9K
Total runs
-54
24-hour runs
289
3-day runs
356
7-day runs
-1.5K
30-day runs

More Information About Kimi-K3-GGUF huggingface.co Model

More Kimi-K3-GGUF license Visit here:

https://choosealicense.com/licenses/kimi-k3

Kimi-K3-GGUF huggingface.co

Kimi-K3-GGUF huggingface.co is an AI model on huggingface.co that provides Kimi-K3-GGUF's model effect (), which can be used instantly with this AtomicChat Kimi-K3-GGUF model. huggingface.co supports a free trial of the Kimi-K3-GGUF model, and also provides paid use of the Kimi-K3-GGUF. Support call Kimi-K3-GGUF model through api, including Node.js, Python, http.

AtomicChat Kimi-K3-GGUF online free

Kimi-K3-GGUF huggingface.co is an online trial and call api platform, which integrates Kimi-K3-GGUF's modeling effects, including api services, and provides a free online trial of Kimi-K3-GGUF, you can try Kimi-K3-GGUF online for free by clicking the link below.

AtomicChat Kimi-K3-GGUF online free url in huggingface.co:

https://huggingface.co/AtomicChat/Kimi-K3-GGUF

Kimi-K3-GGUF install

Kimi-K3-GGUF is an open source model from GitHub that offers a free installation service, and any user can find Kimi-K3-GGUF on GitHub to install. At the same time, huggingface.co provides the effect of Kimi-K3-GGUF install, users can directly use Kimi-K3-GGUF installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

Kimi-K3-GGUF install url in huggingface.co:

https://huggingface.co/AtomicChat/Kimi-K3-GGUF

Url of Kimi-K3-GGUF

Kimi-K3-GGUF huggingface.co Url

Provider of Kimi-K3-GGUF huggingface.co

AtomicChat
ORGANIZATIONS

Other API from AtomicChat

huggingface.co

Total runs: 431
Run Growth: -206
Growth Rate: -52.15%
Updated:July 23 2026
huggingface.co

Total runs: 32
Run Growth: -42
Growth Rate: -135.48%
Updated:July 28 2026