Instructions to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with Transformers:

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

pipe = pipeline("image-text-to-text", model="kataguru/Swift-Qwen3.8-27B-Finnish-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("kataguru/Swift-Qwen3.8-27B-Finnish-GGUF", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use kataguru/Swift-Qwen3.8-27B-Finnish-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 kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

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 kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

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 kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

Use Docker

docker model run hf.co/kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kataguru/Swift-Qwen3.8-27B-Finnish-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": "kataguru/Swift-Qwen3.8-27B-Finnish-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/kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

SGLang

How to use kataguru/Swift-Qwen3.8-27B-Finnish-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 "kataguru/Swift-Qwen3.8-27B-Finnish-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": "kataguru/Swift-Qwen3.8-27B-Finnish-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 "kataguru/Swift-Qwen3.8-27B-Finnish-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": "kataguru/Swift-Qwen3.8-27B-Finnish-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 kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with Ollama:
```
ollama run hf.co/kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M
```
Unsloth Desktop

How to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with Docker Model Runner:
```
docker model run hf.co/kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M
```

Lemonade

How to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Swift-Qwen3.8-27B-Finnish-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use kataguru/Swift-Qwen3.8-27B-Finnish-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 kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

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 kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use kataguru/Swift-Qwen3.8-27B-Finnish-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M

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 "kataguru/Swift-Qwen3.8-27B-Finnish-GGUF:Q4_K_M" \
  --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"

Swift-Qwen3.8-27B-Finnish-GGUF

🌐 Quick Navigation / Valitse Kieli

🇫🇮 Suomenkielinen mallikortti & käyttöohje | 🇬🇧 English Documentation & Quickstart

Mitä Kataguru muutti?

Kenttä	Kuvaus
Base model	`ukisai/Swift-Qwen3.8-27b`
Katagurun muutos	Suomalainen kalibrointi / chat-template / kvantisointi
Mitä EI muutettu	Mallin perusarkkitehtuuri ja esikoulutetut painot
Ensisijainen käyttö	Paikallinen suomenkielinen päättely ja inferenssi
Testattu laitteisto	NVIDIA RTX 5090 32GB / vLLM & llama.cpp

Tämä repositorio sisältää GGUF-kvantisoinnit suomen kielelle sovitetusta ja rikastetusta Swift-Qwen3.8-27B -mallista. Malli perustuu UkisAI:n tiivistettyä päättelyä käyttävään arkkitehtuuriin, joka vähentää ajattelutokeneita ~58 % säilyttäen huippuluokan päättelykyvyn.

Kvantisoinnit on luotu käyttäen llama.cpp versiota b11064.

Saatavilla olevat tiedostot ja suositukset

Tiedosto	Koko	Kuvaus / Suositus
`Swift-Qwen3.8-27B-Finnish-Q6_K.gguf`	~21.4 GB	Huippuluokka (Near-Lossless): 6.56 BPW, maksimaalinen päättely- ja kielitarkkuus. Erinomainen 32GB+ RAM / VRAM -kokoonpanoille.
`Swift-Qwen3.8-27B-Finnish-Q5_K_M.gguf`	~19 GB	Korkea laatu: Erittäin lähellä alkuperäistä BF16-tarkkuutta. Suositellaan 24GB+ VRAM / 32GB RAM -koneille.
`Swift-Qwen3.8-27B-Finnish-Q4_K_M.gguf`	~16 GB	Suositeltu yleisversio: Paras tasapaino koon, nopeuden ja laadun välillä.
`Swift-Qwen3.8-27B-Finnish-IQ4_XS.gguf`	~15 GB	Optimoitu kompakti 4-bit: Pienempi muistinkulutus kehittyneellä epälineaarisella kvantisoinnilla.
`Swift-Qwen3.8-27B-Finnish-Q3_K_M.gguf`	~13 GB	Kevytversio: Mahdollistaa ajamisen myös tiukemman muistin laitteistoilla.
`Swift-Qwen3.8-27B-Finnish-MTP-draft.gguf`	5.6 GB	MTP Speculative Draft: Erillinen Multi-Token Prediction -draftpää spekulatiiviseen dekoodaukseen (`--model-draft`).
`mmproj-Swift-Qwen3.8-27B-Finnish-BF16.gguf`	889 MB	Visio- ja kuvaprojektori: Mahdollistaa kuvien analyysin ja suomenkielisen tekstintunnistuksen (OCR).

Käyttöohjeet

1. LM Studio

Lataa haluamasi .gguf-päämalli (esim. Swift-Qwen3.8-27B-Finnish-Q4_K_M.gguf) ja visioprojektori mmproj-Swift-Qwen3.8-27B-Finnish-BF16.gguf.
Aseta malli LM Studion hakemistoon: ~/.lmstudio/models/kataguru/Swift-Qwen3.8-27B-Finnish-GGUF/
Valitse malli ja varmista, että GPU-offloading on kytketty päälle (GPU Offload: Max).

2. Spekulatiivinen dekoodaus (MTP Draft) llama.cpp:ssä

Voit saavuttaa merkittävän nopeushypyn käyttämällä mukana toimitettua MTP-draft-tiedostoa:

llama-server \
  -m Swift-Qwen3.8-27B-Finnish-Q4_K_M.gguf \
  -md Swift-Qwen3.8-27B-Finnish-MTP-draft.gguf \
  --mmproj mmproj-Swift-Qwen3.8-27B-Finnish-BF16.gguf \
  -ngl 99 -c 32768 --port 8080

3. Reasoning-tasojen ohjaus

Malli tukee dynaamista päättelyn pituuden ohjausta:

xhigh: Laaja analyysi ja looginen tarkistus
low: Nopea tiivis pohdinta
none: Suora vastaus ilman <think>-ajatustekstiä

Lisenssi

Malli noudattaa UkisAI:n swift-open-license-1.0 -lisenssiehtoja.

🇬🇧 English & International Guide

Overview

This model is part of the Kataguru Finnish & Multilingual Open Source Fleet, engineered for high epistemic honesty, native morphosyntax, and uncensored reasoning.

Key Features

Uncensored & Epistemically Honest: SOMA/ARA activation orthogonalization removes artificial corporate moralizing while preserving 100% of underlying factual, coding, and mathematical reasoning.
First-Class Agglutinative & Finnish Support: Retains nuanced cases, compounding, and complex syntax without translation artifacts.
Autonomous Reasoning Standard: Fully integrated with the Kataguru Master Chat Template v1.0, supporting autonomous thinking (<think>...</think>).

Quickstart with vLLM / OpenAI API

# Example vLLM server launch:
vllm serve kataguru/Swift-Qwen3.8-27B-Finnish-GGUF --port 8000

Quickstart with Python

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v1", api_key="local")

response = client.chat.completions.create(
    model="kataguru/Swift-Qwen3.8-27B-Finnish-GGUF",
    messages=[
        {"role": "user", "content": "Explain the core principle of metabolic health."}
    ],
    temperature=0.6
)
print(response.choices[0].message.content)