You can deploy this model and start earning money today!

nightmedia/
Qwen3-42B-A3B-Element6.7-1M-qx86-hi-mlx

Name: Qwen3-42B-A3B-Element6.7-1M-qx86-hi-mlx
Availability: InStock

Views

No views yet

Qwen3-42B-A3B-Element6.7-1M-qx86-hi-mlx

Brainwaves:

brainwave

mxfp4    0.550,0.678,0.868,0.745,0.428,0.798,0.703
qx86-hi  0.559,0.718,0.879,0.757,0.456,0.800,0.722

This is a model merge between

Qwen3-42B-A3B-Element6-1M
Qwen3-42B-A3B-Element7-1M

Showing the qx86-hi performance

brainwave

Element6-1M  0.564,0.714,0.879,0.758,0.452,0.799,0.716
Element7-1M  0.262,0.269,0.381,0.260,0.280,0.515,0.510

It is also the first successful merge of two brainstormed models by DavidAU

Over the parent model, the Element6.7 changed a few arc values, and improved winogrande, without significant loss or gain.

It's just different now.

It repairs the perplexity damage done by brainstorming Element7, creating the first super-MoE: a MoE of two brainstorming MoEs.

It's a MoM: Mixture of MoEs

-G

Use with mlx

pip install mlx-lm

python

from mlx_lm import load, generate

model, tokenizer = load("Qwen3-42B-A3B-Element6.7-1M-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)

nightmedia/Qwen3-42B-A3B-Element6.7-1M-qx86-hi-mlx

Qwen3-42B-A3B-Element6.7-1M-qx86-hi-mlx

Use with mlx

nightmedia/
Qwen3-42B-A3B-Element6.7-1M-qx86-hi-mlx