Rao On-device Base Set
Collection
The base on-device models that make Ambient, Veil, Craft, Ambient Intelligence work. • 7 items • Updated
Mistral-Nemo-Instruct-2407-mlx-4bit
Rao Studios' byte-identical mirror of
mlx-community/Mistral-Nemo-Instruct-2407-4bit
at commit 647ca0751669b21a364c86ccc5df54c4d7e4e91c, which mlx-community converted from
mistralai/Mistral-Nemo-Instruct-2407
with mlx-lm 0.16.0. Every weight and tokenizer file is unchanged; only this card and LICENSE.txt are added.
- Architecture:
MistralForCausalLM, 12B parameters, 40 layers, 128k-vocabulary tokenizer (131072) - Quantization: 4-bit, group size 64
- Weights: about 6.9 GB in two safetensors shards
- License: Apache-2.0 (see
LICENSE.txt)
Mirrored so the models Rao's on-device stack depends on (Ambient, Craft, Veil, via AII) are pinned to a repository Rao controls.
- Downloads last month
- 260
Model size
12B params
Tensor type
U32
·
F16 ·
Hardware compatibility
Log In to add your hardware
4-bit