granite-4.2-3b-q8-hi-mlx

Brainwaves

         arc   arc/e boolq hswag obkqa piqa  wino
q8-hi    0.426,0.570,0.838,0.527,0.410,0.708,0.579
mxfp8    0.393,0.486,0.739,0.492,0.394,0.724,0.603
Thinking
mxfp8    0.344,0.380,0.622

Previous model

granite-4.1-3b

         arc   arc/e boolq hswag obkqa piqa  wino
mxfp8    0.406,0.581,0.821,0.484,0.434,0.712,0.559

The performance in mxfp8 dips significantly below the old base. It looks like this model does better with int quants.

More metrics coming soon.

-G

---

This model granite-4.2-3b-q8-hi-mlx was converted to MLX format from ibm-granite/granite-4.2-3b using mlx-lm version 0.31.3.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("granite-4.2-3b-q8-hi-mlx")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True, return_dict=False,
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)