Downloads · 30 days
29
4% of all-time downloads
Qwen3.6-35B-A3B-BF16-MLX
Vaynel/Qwen3.6-35B-A3B-BF16-MLX
Qwen3.6-35B-A3B-BF16-MLX is a machine learning model from Vaynel. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as apache-2.0.
This model was converted to MLX format from Qwen/Qwen3.6-35B-A3B using mlx-lm.
All-time downloads
723
Public
Parameters
34.7B
69.3 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors69.3 GB · 100%
At a glance
- Library
- mlx
- License
- apache-2.0
- Model type
- qwen3_5_moe
- Access
- Public
- Created
- Apr 24, 2026
- Updated
- Apr 24, 2026
- SHA
- bf57ea33
Base models
Model card
From the Hugging Face model README
Vaynel/Qwen3.6-35B-A3B-BF16-MLX
This model was converted to MLX format from Qwen/Qwen3.6-35B-A3B using mlx-lm.
Full BF16 precision — no quantization. For the 4-bit quantized version see mlx-community/Qwen3.6-35B-A3B-4bit.
Benchmarks (M3 Ultra, 512GB)
| Quantization | c=1 tok/s | c=3 tok/s (per-req) | c=3 agg tok/s |
|---|---|---|---|
| 4-bit MLX | 90.8 | 37.7 | 162.7 |
| BF16 MLX (this) | 50.3 | 27.7 | 118.2 |
BF16 is ~44% slower per request due to higher memory bandwidth requirements (~65GB vs 19GB). Use 4-bit for throughput; use BF16 for maximum output quality on reasoning tasks.
Usage
pip install mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("Vaynel/Qwen3.6-35B-A3B-BF16-MLX")
response = generate(model, tokenizer, prompt="Write a binary search in Python.", max_tokens=512)
print(response)
Model Details
- Architecture: Qwen3MoeForCausalLM (MoE, 35B total / 3.6B active)
- Context: 131,072 tokens
- Precision: BF16 (bfloat16)
- Size: ~65GB
- Hardware tested: Apple M3 Ultra 512GB
- Library
- mlx
- Type
- qwen3_5_moe
- License
- apache-2.0
- Languages
- en
- Created
- Apr 24, 2026
- Updated
- Apr 24, 2026