Downloads · 30 days
10
3% of all-time downloads
zaydiscold/Qwopus3.6-35B-A3B-v1-MLX-3bit
Qwopus3.6-35B-A3B-v1-MLX-3bit is a machine learning model from zaydiscold. 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 other.
MLX 3-bit conversion of Jackrong/Qwopus3.6-35B-A3B-v1.
Downloads · 30 days
10
3% of all-time downloads
All-time downloads
336
Public
Parameters
34.7B
15.2 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors15.2 GB · 100%
How the weights are stored.
U3234.7B · 100%
From the Hugging Face model README
MLX 3-bit conversion of Jackrong/Qwopus3.6-35B-A3B-v1.
Converted directly from the original HF bf16 safetensors. Not from GGUF, not chained from another quant.
| Variant | Repo | Disk | ~Min unified RAM | Role |
|---|---|---|---|---|
| MLX bf16 | Qwopus3.6-35B-A3B-v1-MLX-bf16 | 69.3 GB | ~72 GB | Reference |
| MLX 8bit | Qwopus3.6-35B-A3B-v1-MLX-8bit | 36.8 GB | ~40 GB | Near-lossless |
| MLX 6bit | Qwopus3.6-35B-A3B-v1-MLX-6bit | 28.2 GB | ~32 GB | Quality / size middle |
| MLX 4bit | Qwopus3.6-35B-A3B-v1-MLX-4bit | 19.5 GB | ~22 GB | Standard daily-use tier |
| MLX 3bit (this repo) | this | 15.2 GB | ~18 GB | Smallest practical |
Collection: Qwopus3.6-35B-A3B-v1 MLX
pip install mlx-lm
mlx_lm.generate --model zaydiscold/Qwopus3.6-35B-A3B-v1-MLX-3bit \
--prompt "Explain quantum entanglement in one paragraph" --max-tokens 200
python -m mlx_lm convert \
--hf-path Jackrong/Qwopus3.6-35B-A3B-v1 \
--mlx-path ./Qwopus3.6-35B-A3B-v1-MLX-3bit \
-q --q-bits 3
Q4_K_M is a llama.cpp format. MLX has no literal Q4_K_M — MLX 4-bit is the practical peer at a different quantizer.