Instructions to use unsloth/Kimi-K3-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use unsloth/Kimi-K3-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="unsloth/Kimi-K3-GGUF")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("unsloth/Kimi-K3-GGUF", device_map="auto")

llama-cpp-python

How to use unsloth/Kimi-K3-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="unsloth/Kimi-K3-GGUF",
	filename="UD-IQ1_M/Kimi-K3-UD-IQ1_M-00001-of-00016.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": [
				{
					"type": "text",
					"text": "Describe this image in one sentence."
				},
				{
					"type": "image_url",
					"image_url": {
						"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
					}
				}
			]
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use unsloth/Kimi-K3-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
llama cli -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
llama cli -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
./llama-cli -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
./build/bin/llama-cli -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Use Docker

docker model run hf.co/unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

LM Studio
Jan

vLLM

How to use unsloth/Kimi-K3-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "unsloth/Kimi-K3-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Kimi-K3-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

SGLang

How to use unsloth/Kimi-K3-GGUF with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "unsloth/Kimi-K3-GGUF" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Kimi-K3-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "unsloth/Kimi-K3-GGUF" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Kimi-K3-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Ollama
How to use unsloth/Kimi-K3-GGUF with Ollama:
```
ollama run hf.co/unsloth/Kimi-K3-GGUF:UD-Q4_K_XL
```

Unsloth Studio

How to use unsloth/Kimi-K3-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for unsloth/Kimi-K3-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for unsloth/Kimi-K3-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for unsloth/Kimi-K3-GGUF to start chatting

How to use unsloth/Kimi-K3-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "unsloth/Kimi-K3-GGUF:UD-Q4_K_XL"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use unsloth/Kimi-K3-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use unsloth/Kimi-K3-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "unsloth/Kimi-K3-GGUF:UD-Q4_K_XL" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use unsloth/Kimi-K3-GGUF with Docker Model Runner:
```
docker model run hf.co/unsloth/Kimi-K3-GGUF:UD-Q4_K_XL
```

Lemonade

How to use unsloth/Kimi-K3-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull unsloth/Kimi-K3-GGUF:UD-Q4_K_XL

Run and chat with the model

lemonade run user.Kimi-K3-GGUF-UD-Q4_K_XL

List all available models

lemonade list

THERES NO GGUFS

by GGUFGuy - opened 1 day ago

Discussion

GGUFGuy

1 day ago

THERES NO GGUFS

Ognjen813

1 day ago

WHERE THEM GGUFS AT!?

GGUFGuy

1 day ago

THEY LEFT THIS MODEL AND WENT TO THE MOON

danielhanchen

Unsloth AI org about 23 hours ago

We are working on them, please be patient. The model is HUGE!! :)

d4nnyy

about 22 hours ago

@danielhanchen yo! we're on it too at atomic chat :))) q2_k is uploading rn

simonko912

about 22 hours ago

great! I cant wait for ggufs of a model that i will probably not be able to run

allarac

about 22 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

simonko912

about 22 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

My friend can somehow run up to 2T, while im stuck at home with around 20b max

OliverBright155

about 21 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

Me personally I'm splitting the og files across multiple NVMe drives and loading only those experts which are needed into the 128gb memory of a secondhand server I got off ebay, as I'm working on making it more efficient I'm beginning to get 1+t/s on CPU-only inference :D

dk38

about 20 hours ago

We are working on them, please be patient. The model is HUGE!! :)

LMAO ;)

simonko912

about 20 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

Me personally I'm splitting the og files across multiple NVMe drives and loading only those experts which are needed into the 128gb memory of a secondhand server I got off ebay, as I'm working on making it more efficient I'm beginning to get 1+t/s on CPU-only inference :D

I have like 64 cores of a unknown cpu and 2x 4090s, so that wouldnt fit. Maybe on cpu since i have 256gb ddr5 ecc

allarac

about 20 hours ago

•

edited about 20 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

Me personally I'm splitting the og files across multiple NVMe drives and loading only those experts which are needed into the 128gb memory of a secondhand server I got off ebay, as I'm working on making it more efficient I'm beginning to get 1+t/s on CPU-only inference :D

I have like 64 cores of a unknown cpu and 2x 4090s, so that wouldnt fit. Maybe on cpu since i have 256gb ddr5 ecc

Doing [insert weird stuff here] to get 1 tps out of it, is not really useful at all anyway. The only way to change that if everyone actually STOPPED buying overpriced hardware with little capacity.

simonko912

about 20 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

Me personally I'm splitting the og files across multiple NVMe drives and loading only those experts which are needed into the 128gb memory of a secondhand server I got off ebay, as I'm working on making it more efficient I'm beginning to get 1+t/s on CPU-only inference :D

I have like 64 cores of a unknown cpu and 2x 4090s, so that wouldnt fit. Maybe on cpu since i have 256gb ddr5 ecc

Doing [insert weird stuff here] to get 1 tps out of it, is not really useful at all anyway. The only way to change that if everyone actually STOPPED buying overpriced hardware with little capacity.

I personally think 0.5tok/s is the minimum i like

worthant

about 18 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

Me personally I'm splitting the og files across multiple NVMe drives and loading only those experts which are needed into the 128gb memory of a secondhand server I got off ebay, as I'm working on making it more efficient I'm beginning to get 1+t/s on CPU-only inference :D

I have like 64 cores of a unknown cpu and 2x 4090s, so that wouldnt fit. Maybe on cpu since i have 256gb ddr5 ecc

Sadly smallest IQ1_S quant won't be less than around 500 GB 😢
Training 1bit and 2bit quants right know and they will be 500-800GB, i can tell you that much

GGUFGuy

about 13 hours ago

We are working on them, please be patient. The model is HUGE!! :)

ok unsloth

simonko912

about 11 hours ago

Yes it is truly awesome that these weights are available and that unsloth offers great quantization solutions. However, who on earth is supposed to actually run this mega model. The hardware simply does only exist in datacenters, and there it is prohibitively expensive compared to LLM api providers.
I am wondering, why is anyone "waiting" for these weights?

Me personally I'm splitting the og files across multiple NVMe drives and loading only those experts which are needed into the 128gb memory of a secondhand server I got off ebay, as I'm working on making it more efficient I'm beginning to get 1+t/s on CPU-only inference :D

I have like 64 cores of a unknown cpu and 2x 4090s, so that wouldnt fit. Maybe on cpu since i have 256gb ddr5 ecc

Sadly smallest IQ1_S quant won't be less than around 500 GB 😢
Training 1bit and 2bit quants right know and they will be 500-800GB, i can tell you that much

Okay i didnt expect him to be able to run it

d4nnyy

about 10 hours ago

•

edited about 10 hours ago

if someone needs it - you can try these quants, interesting how it turned out - https://huggingface.co/AtomicChat/Kimi-K3-GGUF

IQ1_S - 555 GB

on the 5 dgx sparks for example 🤪

GGUFGuy

about 10 hours ago

finally, they uploaded the ggufs
but https://huggingface.co/AtomicChat/Kimi-K3-GGUF is better and has smaller quants like IQ1_S

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment