Downloads · 30 days
234
36% of all-time downloads
majentik/Mistral-Medium-3.5-128B-MLX-2bit
Mistral-Medium-3.5-128B-MLX-2bit is a image-text-to-text model from majentik. Use it for the image-text-to-text 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 2bit (affine, group size 32) quantized variant of mistralai/Mistral-Medium-3.5-128B (text tower quantized; vision tower and projector retained in BF16; source checkpoint is FP8) for Apple silicon via mlx-lm.
Downloads · 30 days
234
36% of all-time downloads
All-time downloads
648
Public
Parameters
128B
54.9 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors54.9 GB · 100%
How the weights are stored.
U32123B · 97%
From the Hugging Face model README
MLX 2bit (affine, group size 32) quantized variant of mistralai/Mistral-Medium-3.5-128B (text tower quantized; vision tower and projector retained in BF16; source checkpoint is FP8) for Apple silicon via mlx-lm.
22b2b868a15677cfa6061277ed2f653d1349a9ab (Modified MIT — revenue carve-out below $20M/month).mlx_lm.convert (mlx-lm 0.31.3): affine,
2-bit, group size 32.Before upload this pack passed a deterministic coherence gate: greedy
48-token chat generation loaded through
mlx_lm.load, judged for emptiness, repetition loops, multi-script
gibberish, and special-token debris. Verdict: ok.
pip install mlx-lm
mlx_lm.generate --model majentik/Mistral-Medium-3.5-128B-MLX-2bit --prompt "Hello"
Modified MIT (upstream LICENSE): free use with a revenue carve-out — organisations
above $20M/month revenue need a commercial agreement with Mistral AI. This is not a
plain MIT license; read the upstream terms before commercial deployment.
| Benchmark | Score |
|---|---|
| arc_easy_acc | 0.7400 |
| hellaswag_acc | 0.5150 |