Instructions to use Shadow0482/tb-x with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Shadow0482/tb-x with Transformers:

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

pipe = pipeline("text-generation", model="Shadow0482/tb-x")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Shadow0482/tb-x")
model = AutoModelForCausalLM.from_pretrained("Shadow0482/tb-x", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use Shadow0482/tb-x 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 Shadow0482/tb-x:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Shadow0482/tb-x:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Shadow0482/tb-x:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Shadow0482/tb-x:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf Shadow0482/tb-x:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf Shadow0482/tb-x:Q4_K_M

Use Docker

docker model run hf.co/Shadow0482/tb-x:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use Shadow0482/tb-x with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Shadow0482/tb-x"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Shadow0482/tb-x",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Shadow0482/tb-x:Q4_K_M

SGLang

How to use Shadow0482/tb-x 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 "Shadow0482/tb-x" \
    --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": "Shadow0482/tb-x",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "Shadow0482/tb-x" \
        --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": "Shadow0482/tb-x",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Shadow0482/tb-x with Ollama:
```
ollama run hf.co/Shadow0482/tb-x:Q4_K_M
```
Unsloth Desktop

How to use Shadow0482/tb-x with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Shadow0482/tb-x: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": "Shadow0482/tb-x:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Shadow0482/tb-x with Docker Model Runner:
```
docker model run hf.co/Shadow0482/tb-x:Q4_K_M
```

Lemonade

How to use Shadow0482/tb-x with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Shadow0482/tb-x:Q4_K_M

Run and chat with the model

lemonade run user.tb-x-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use Shadow0482/tb-x with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Shadow0482/tb-x: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 Shadow0482/tb-x:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Shadow0482/tb-x with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Shadow0482/tb-x: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 "Shadow0482/tb-x: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"

tb-x

File size: 4,731 Bytes

---
tags:
  - gguf
  - llama.cpp
  - llama-cpp
  - unsloth
  - text-generation
  - transformers
  - qwen2
  - agentic-workflows
  - tool-calling
license: apache-2.0
language:
  - en
base_model: orangejuicesmith/Mythos-Nano-safetensor
---

# TB-X — Mythos Nano GGUF

**TB-X** is a GGUF conversion of [Mythos Nano](https://huggingface.co/orangejuicesmith/Mythos-Nano-safetensor), prepared using [Unsloth](https://github.com/unslothai/unsloth) and made compatible with the [llama.cpp](https://github.com/ggml-org/llama.cpp) inference ecosystem.

The repository provides multiple GGUF precision and quantization variants, allowing users to choose an appropriate balance between model size, memory usage, inference performance, and output quality.

## Model Information

| Property | Details |
|---|---|
| Repository | `Shadow0482/tb-x` |
| Model name | TB-X |
| Base model | `orangejuicesmith/Mythos-Nano-safetensor` |
| Architecture tag | Qwen2 |
| Format | GGUF |
| Conversion | Unsloth |
| Inference engine | llama.cpp |
| Language | English |
| License | Apache 2.0 |

## Development & Agent Workflow

TB-X was developed with a teacher-guided, tool-driven agent workflow.

The development process used:

- **Claude Opus 5.5** — Teacher model.
- **GPT-6 Astra** — Teacher and reasoning support.
- **Tool orchestration** — Tool calls and agent interactions used throughout the workflow.

The workflow involved approximately **500 million tool-call and agent-flow interactions**.

This figure describes the reported development workflow activity; it does not represent the model's parameter count, training tokens, or number of training samples.

> The teacher models were used as part of the development workflow. They are not inference dependencies of the released GGUF files.

## Available Model Files

TB-X includes the following GGUF variants:

| File | Precision / Quantization |
|---|---|
| `Mythos-Nano-safetensor.BF16.gguf` | BF16 |
| `Mythos-Nano-safetensor.F16.gguf` | F16 |
| `Mythos-Nano-safetensor.Q2_K_L.gguf` | Q2_K_L |
| `Mythos-Nano-safetensor.Q3_K_M.gguf` | Q3_K_M |
| `Mythos-Nano-safetensor.Q4_0.gguf` | Q4_0 |
| `Mythos-Nano-safetensor.Q4_1.gguf` | Q4_1 |
| `Mythos-Nano-safetensor.Q4_K_M.gguf` | Q4_K_M |
| `Mythos-Nano-safetensor.Q5_K_M.gguf` | Q5_K_M |
| `Mythos-Nano-safetensor.Q6_K.gguf` | Q6_K |
| `Mythos-Nano-safetensor.Q8_0.gguf` | Q8_0 |

### Choosing a Quantization

- **BF16 / F16:** Higher precision, with greater memory requirements.
- **Q8_0:** High-quality quantized option.
- **Q6_K / Q5_K_M:** Good options when prioritizing quality with reduced memory usage.
- **Q4_K_M / Q4_0 / Q4_1:** Practical lower-memory options.
- **Q3_K_M / Q2_K_L:** More aggressive compression for constrained hardware.

Actual memory requirements depend on the architecture, context length, runtime, and inference configuration.

## Usage with llama.cpp

Install or build a recent version of [llama.cpp](https://github.com/ggml-org/llama.cpp).

### 1. Text Generation

Run TB-X using the Hugging Face model repository:

```bash
llama-cli -hf Shadow0482/tb-x --jinja
```

### 2. Chat Template

The `--jinja` option enables llama.cpp's Jinja chat-template handling.

This is useful when the model repository provides a compatible chat template.

### 3. Run a Local GGUF File

Download your preferred quantization and run it directly:

```bash
llama-cli \
  -m Mythos-Nano-safetensor.Q4_K_M.gguf \
  --jinja \
  -cnv
```

Replace the filename with your selected GGUF variant.

### 4. OpenAI-Compatible API Server

Start a local inference server:

```bash
llama-server \
  -hf Shadow0482/tb-x \
  --jinja \
  --host 0.0.0.0 \
  --port 8080
```

The API is available at:

```text
http://localhost:8080/v1
```

You can connect compatible applications and clients to the local server.

## Compatibility

TB-X is designed for use with the llama.cpp ecosystem.

- llama-cli
- llama-server
- OpenAI-compatible API clients
- Compatible GGUF applications

Tool calling and agent functionality depend on the model's capabilities, chat template, and the runtime's tool-calling implementation. A GGUF conversion alone does not guarantee native function-calling support.

## Credits & Acknowledgements

- **Base model:** [orangejuicesmith/Mythos-Nano-safetensor](https://huggingface.co/orangejuicesmith/Mythos-Nano-safetensor)
- **Conversion framework:** [Unsloth](https://github.com/unslothai/unsloth)
- **Inference engine:** [llama.cpp](https://github.com/ggml-org/llama.cpp)
- **Teacher models:** Claude Opus 5.5 and GPT-6 Astra

## License

The base model metadata specifies the Apache 2.0 license. Please review the base model's license and applicable conditions before redistributing or using derivative artifacts.