Instructions to use beyoru/Orbit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use beyoru/Orbit with Transformers:

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

pipe = pipeline("text-generation", model="beyoru/Orbit")
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 AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("beyoru/Orbit")
model = AutoModelForMultimodalLM.from_pretrained("beyoru/Orbit", device_map="auto")
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?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use beyoru/Orbit with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/beyoru/Orbit

SGLang

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

Docker Model Runner
How to use beyoru/Orbit with Docker Model Runner:
```
docker model run hf.co/beyoru/Orbit
```

Orbit

File size: 2,318 Bytes

---
license: mit
language:
- vi
- en
base_model: Qwen/Qwen3.5-35B-A3B
library_name: transformers
pipeline_tag: text-generation
tags:
- agent
- tool-use
- reasoning
- esft
- claude-opus-5
- xhigh
- distillation
- coding
- terminal
- moe
datasets:
- beyoru/claude-opus-5-xhigh-workload-agent
---

# Orbit

Expert-Specialized Fine-Tune (ESFT) of
[`Qwen/Qwen3.5-35B-A3B`](https://huggingface.co/Qwen/Qwen3.5-35B-A3B)
for Vietnamese multi-turn tool-use, trained on reasoning traces.

Architecture is unchanged from the base model.

This model was training on the distillation dataset from `claude-opus-5`, effort `xhigh` for my custome workflow

# Capabilities
## Multi-turn tool use
This model is trained to maintain context across multiple tool interactions,
rather than treating each tool call as an isolated operation.
This makes it suitable for workflows where the result of one action determines
the next action.

## Tool selection

The model is trained on trajectories containing tool selection and execution,
allowing it to reason about:

- which tool should be used
- when a tool call is necessary
- what arguments should be provided
- how to interpret tool results
- whether additional actions are required

The training data contains high-effort reasoning trajectorie

## Training

- **ESFT**: only selected MoE experts are trained (router frozen).
- trainable: ~0.94B of 35.6B parameters (2.6%)
- expert selection: `top_p = 0.20`, ~7.5 of 256 experts per layer
- single NVIDIA GB10 (121 GB unified memory)

## Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

tk = AutoTokenizer.from_pretrained("beyoru/Clawd-Agent")
model = AutoModelForCausalLM.from_pretrained("beyoru/Clawd-Agent", dtype="auto", device_map="auto")

msgs = [{"role": "user", "content": "..."}]
ids = tk.apply_chat_template(msgs, tools=TOOLS, add_generation_prompt=True, return_tensors="pt")
```

Serving with vLLM:

```bash
vllm serve beyoru/Orbit --max-model-len 8192
```

The chat template emits `<think>\n` in the generation prompt, so the model continues the
reasoning block and closes it with `</think>` before the answer.

## Note

Inherits the base model's MIT license. Fine-tuned on a narrow task distribution — evaluate on
your own workload before relying on it for anything outside multi-turn tool use.