Qwen3-1.7B-Nyx-dq48-mlx

Performance evaluation

arc_challenge
acc 0.329, norm 0.340, stderr 0.013
arc_easy
acc 0.437, norm 0.381, stderr 0.009
boolq
acc 0.621, norm 0.621, stderr 0.008
hellaswag
acc 0.427, norm 0.507, stderr 0.004
openbookqa
acc 0.212, norm 0.370, stderr 0.021
piqa
acc 0.693, norm 0.674, stderr 0.010
winogrande
acc 0.548, norm 0.548, stderr 0.013

This model Qwen3-1.7B-Nyx-dq48-mlx was converted to MLX format from bunnycore/Qwen3-1.7B-Nyx using mlx-lm version 0.26.0.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("Qwen3-1.7B-Nyx-dq48-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
    )

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