---
pipeline_tag: text-generation
library_name: mlx
model_name: K2-Horizon-7B
language:
- en
license: apache-2.0
datasets:
- IFM/K2-Horizon-Pretrain-Data
- IFM/K2-Horizon-Midtrain-Data
tags:
- k2-horizon
- 7b
- dense
- open-weights
- ifm
- mlx
- mlxhub
base_model: IFM/K2-Horizon-7B
---

[![Open in MLXHub](https://mlxhub.app/assets/badge-open-in-mlxhub.svg)](https://mlxhub.app/open-model?repo=DreamFoundries/K2-Horizon-7B-MLX-4bit)

# K2 Horizon 7B MLX 4-bit

MLX conversion of [IFM/K2-Horizon-7B](https://huggingface.co/IFM/K2-Horizon-7B), created specifically for MLXHub with the DreamFoundries `mlx-lm` fork at `0f74c0e`. Affine 4-bit quantization uses group size 64. The K2 routers (`mlp.gate` and, where present, `self_attn.v_router`) remain unquantized by the model implementation.

The original model is by [Institute of Foundation Models (IFM)](https://huggingface.co/IFM) and is released under Apache-2.0. Comparative quality and performance benchmarks are not available for this conversion.

## Use with MLX

```python
from mlx_lm import load, generate
model, tokenizer = load("DreamFoundries/K2-Horizon-7B-MLX-4bit")
response = generate(model, tokenizer, prompt="hello", verbose=True)
```

---

[![Download MLXHub](https://mlxhub.app/assets/badge-download-mlxhub.svg)](https://apps.apple.com/app/apple-store/id6766485144?pt=121945436&ct=HuggingFace&mt=8)