gemma-3-12b-it-vl-GLM-4.7-Flash-Polaris-Heretic-Uncensored-Thinking-qx86-hi-mlx

This is a nuslerp merge of:

• DavidAU/gemma-3-12b-it-vl-GLM-4.7-Flash-Heretic-Uncensored-Thinking
• DavidAU/gemma-3-12b-it-vl-Polaris-Heretic-Uncensored-Thinking

Brainwaves

          arc   arc/e boolq hswag obkqa piqa  wino
qx86-hi   0.607,0.787,0.870,0.720,0.484,0.797,0.708
qx64-hi   0.586,0.778,0.867,0.717,0.466,0.798,0.710
mxfp4     0.561,0.749,0.857,0.712,0.432,0.781,0.709

gemma-3-27b-it-heretic
q8        0.557,0.711,0.868,0.533,0.452,0.706,0.695

gemma-3-12b-it-vl-GLM-4.7-Flash-Heretic-Uncensored-Thinking
qx86-hi   0.585,0.756,0.874,0.724,0.462,0.798,0.717

gemma-3-12b-it-vl-Polaris-Heretic-Uncensored-Thinking
qx86-hi   0.619,0.791,0.859,0.705,0.482,0.765,0.714

Perplexity:
glm-4.7   12.799 ± 0.128
polaris   14.909 ± 0.155

merged    10.553 ± 0.101 

-G

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("gemma-3-12b-it-vl-GLM-4.7-Flash-Polaris-Heretic-Uncensored-Thinking-qx86-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)