Catalog
Our AI Models
All models are released open source on Hugging Face with a variety of architectures and training objectives.
Orbit 3 8B Llama Thinking
Text Generation
A general-purpose text generation model with 8 billion parameters at FP16 precision, built for broad multilingual tasks with enhanced reasoning on cybersecurity problems.
Precision
FP16
Highlights
General-purpose · Cyber reasoning
Orbit 3 Nano
Text Generation
An improved model in Indonesian language understanding and reasoning, trained using 86,000 high-quality data samples.
Precision
BF16
Highlights
General-purpose · Indonesian
Orbit 3.2 Nano
Text Generation
An enhanced Orbit 3 Nano model for Indonesian language understanding, trained using 15,000 high-quality data samples.
Precision
BF16
Highlights
General-purpose · Indonesian
Orbit 3.5 Nano
Text Generation
Orbit 3.5 Nano is a development of the Orbit 3.2 Nano model, which has improved its ability to understand more complex Indonesian language contexts.
Precision
BF16
Highlights
General-purpose · Indonesian
Orbit 4 Gemma E4B it
Text Generation
The latest model from the Orbit 4 series trained on a large-scale, high-quality Indonesian language dataset to improve understanding of more complex Indonesian languages compared to the previous Orbit 3 series.
Precision
F16
Highlights
General-purpose · Indonesian
Technology
Technology we use
We build and train our models with the best open source stack in the industry.

NVIDIA
Transformers

PyTorch
Hugging Face

CUDA

ONNX

Docker

Kubernetes