baa-ai/Qwen3.6-35B-A3B-RAM-25GB-MLX

Qwen3.6-35B-A3B — 25GB (MLX)

Mixed-precision MLX build of Qwen/Qwen3.6-35B-A3B, prepared by baa.ai.

Built at the predicted global (quality max) operating point.

Metrics

Metric	Value
In-memory footprint	~25 GiB
Size on disk	26.3 GB
Average bits per weight	5.29
Group size	64
Framework	MLX (Apple Silicon)
Source	Qwen/Qwen3.6-35B-A3B (BF16, 71.9 GB)

Benchmarks

Benchmarks pending.

Recommended inference settings

sampler_params = {
    "temperature": 1.0,
    "top_p": 0.95,
    "top_k": 40,
    "repetition_penalty": 1.1,
    "max_tokens": 8192,
}

Usage

from mlx_lm import load, generate
from mlx_lm.sample_utils import make_sampler, make_logits_processors

model, tokenizer = load("baa-ai/Qwen3.6-35B-A3B-RAM-25GB-MLX")

sampler = make_sampler(temp=1.0, top_p=0.95, top_k=40)
logits_processors = make_logits_processors(repetition_penalty=1.1)

prompt = tokenizer.apply_chat_template(
    [{"role": "user", "content": "Write a Python function that reverses a string."}],
    tokenize=False,
    add_generation_prompt=True,
)

response = generate(model, tokenizer, prompt=prompt, max_tokens=8192,
                    sampler=sampler, logits_processors=logits_processors)
print(response)

Hardware

• Apple Silicon Mac with ~32 GB unified memory recommended.

Variants

Variant	Size	Link
19 GB	20.6 GB	baa-ai/Qwen3.6-35B-A3B-RAM-19GB-MLX
25 GB	26.3 GB	baa-ai/Qwen3.6-35B-A3B-RAM-25GB-MLX

License

Apache 2.0 — inherited from Qwen/Qwen3.6-35B-A3B.

---

Quantized by baa.ai