# What is Halo?

Halo is White Circle's open-source framework for post-training open models. It is built on the Hugging Face stack and adds the parallelism that large and mixture-of-experts models need, from one GPU to many nodes.

## What it is

Halo takes a Hugging Face model and a dataset, trains, and writes a Hugging Face model back, loadable with `from_pretrained`. It is built on Transformers, [TRL](/glossary/trl) and Accelerate, and a run is one YAML config.

## What it adds

- **Parallelism.** Expert, context, tensor and expert-tensor parallelism on native Hugging Face models, so a large or [mixture-of-experts](/glossary/mixture-of-experts) model can be split across GPUs.
- **Models.** Dense models and fifteen mixture-of-experts families. Ten have a cookbook.
- **Methods.** [Supervised fine-tuning](/glossary/supervised-fine-tuning), preference tuning, reward modeling, classification, distillation, embedding models, and [GRPO](/glossary/grpo): offline, online, and asynchronous with multi-turn environments.

In the [Halo tech report](https://whitecircle.com/research/halo), on gpt-oss-20b with 4,096-token sequences at batch 1 on B300 GPUs, Halo ran at 20,786 tokens per second per GPU, about 4.8 times stock TRL's 4,320.

## How you run it

Halo runs inside a Docker image: `halo:blackwell` for NVIDIA B200 and B300 GPUs, `halo:hopper` for H100 and H200. One command starts a run:

```bash
halo launch sft examples/sft/qwen3/qwen3-4b-ultrachat.yaml -n 8
```

`sft` is the method and `-n 8` uses eight GPUs. Start with the [quickstart](/halo/docs/quickstart), read the [docs](/halo/docs), or see the [Halo page](/halo). The [license](/glossary/is-halo-open-source) has its own page.

