Build with our open-weight AI model, Orbit38BLlamaThinking.
from transformers import AutoTokenizer, AutoModelForCausalLM
# Load model directly from Hugging Face
model_id = "OpenPathAI/Orbit-3-8B-Llama-thinking"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id,
torch_dtype="float16",
device_map="auto")Downloading model.safetensors: 100% |██████████| 16.0G/16.0G
Loading weights: 100% |██████████| 291/291
✓ Model loaded — Orbit-3-8B-Llama-thinking
About Us
Open, multilingual AI for a global audience
OpenPathAI builds large language, vision, and generative models designed for a global audience. Every model is released publicly on Hugging Face with an open license for research and commercial use.
Our focus: transparent, efficient, and multilingual AI that performs consistently across languages — no region-specific bias, no black boxes.
4
Models Released
Open
Community
MIT
License
∞
Free to Use
Highlights
Built for production
Every model is engineered for quality, speed, and openness — ready to ship from prototype to production scale.
Fast Inference
Models optimized for low latency on consumer hardware and production servers alike.
Open License
Released under MIT and Apache 2.0. Free to use for research and commercial work.
Wide Compatibility
Supports PyTorch, ONNX, TensorRT, and llama.cpp for flexible deployment.
Full Documentation
Every model ships with a model card, code samples, and usage guides.
Ready to try our models?
All models are free on Hugging Face. Just a few lines of code to get started.