Instructions to use thetinkerer/Qwen3.8-27B-TOS-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use thetinkerer/Qwen3.8-27B-TOS-v1 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 thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf thetinkerer/Qwen3.8-27B-TOS-v1: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 thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf thetinkerer/Qwen3.8-27B-TOS-v1: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 thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M

Use Docker

docker model run hf.co/thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M

LM Studio
Jan
Ollama
How to use thetinkerer/Qwen3.8-27B-TOS-v1 with Ollama:
```
ollama run hf.co/thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M
```
Unsloth Desktop

How to use thetinkerer/Qwen3.8-27B-TOS-v1 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf thetinkerer/Qwen3.8-27B-TOS-v1: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": "thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use thetinkerer/Qwen3.8-27B-TOS-v1 with Docker Model Runner:
```
docker model run hf.co/thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M
```

Lemonade

How to use thetinkerer/Qwen3.8-27B-TOS-v1 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.8-27B-TOS-v1-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use thetinkerer/Qwen3.8-27B-TOS-v1 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf thetinkerer/Qwen3.8-27B-TOS-v1: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 thetinkerer/Qwen3.8-27B-TOS-v1:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use thetinkerer/Qwen3.8-27B-TOS-v1 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf thetinkerer/Qwen3.8-27B-TOS-v1: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 "thetinkerer/Qwen3.8-27B-TOS-v1: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"

Qwen3.8 27B Instruct finetuned and modifed for HolyC/TempleOS API knowledge

This is an alpha release v1, your mileage will be limited. The purpose is HolyC coding and code completion, but some general knowledge questions may also yield results. Obviously the amount of TempleOS related training data is severly limited in scope so don't expect miracles. This is trained on only public domain TempleOS code and not any 3rd party or TinkerOS/ZealOS/etc code. This model does still support thinking and tool usage so it should be suitable to use inside your IDE to read/write/modify HolyC files. Some non-English tokens and vision capabilities have been removed to help remove model size and some HolyC/TempleOS specific tokens have been added in their place to improve performance.

LM Studio setup

Copy the GGUF files to a directory structure like:
- Linux: /home/youruser/.lmstudio/models/templeos/qwen38_TOS_v1_Q3_K_M/qwen38-27b-instruct-v1-Q3_K_M.gguf
- Windows: C:\Users\[Your Username]\LM Studio\models\templeos\qwen38_TOS_v1_Q3_K_M\qwen38-27b-instruct-v1-Q3_K_M.gguf
Setup the System Prompt using the recommended prompt from below in the model Inference tab System Prompt:
Set a reasoning limit so it doesn't get stuck thinking forever about things like the differences between C and HolyC

Recommended minimal coding system prompt:

You are answering questions about HolyC, the programming language of Terry Davis's TempleOS operating system. Use real TempleOS API names (MAlloc/Free, StrCpy/StrLen, FileRead/FileWrite, ...). Never use any of the following: enums, function macros, prefix-casting, C-standard-library names, typedef, ternary operators, or continue.

Say a prayer, try a question (and maybe profit?):

Note: lower quantizations like Q3 will run faster and use less VRAM, however the output quality likely will be worse than Q4/Q5

See this chart to get an idea how quantization effects performance and VRAM requirements:

Original chart source: https://huggingface.co/unsloth/Qwen3.8-27B-GGUF/blob/main/README.md

Downloads last month: 167

GGUF

Model size

26B params

Architecture

qwen35

Hardware compatibility

3-bit

4-bit

5-bit

6-bit

8-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for thetinkerer/Qwen3.8-27B-TOS-v1

Base model

Qwen/Qwen3.8-27B

Quantized

(1165)

this model

thetinkerer
/

Qwen3.8-27B-TOS-v1

Qwen3.8 27B Instruct finetuned and modifed for HolyC/TempleOS API knowledge

LM Studio setup

Recommended minimal coding system prompt:

Recommended minimal knowledge question system prompt:

Say a prayer, try a question (and maybe profit?):

Note: lower quantizations like Q3 will run faster and use less VRAM, however the output quality likely will be worse than Q4/Q5

Model tree for thetinkerer/Qwen3.8-27B-TOS-v1