Downloads · 30 days
26
10% of all-time downloads
Ministral-3-3B-Instruct-2512-4bit
Koutouzov1973/Ministral-3-3B-Instruct-2512-4bit
Ministral-3-3B-Instruct-2512-4bit is a text generation model from Koutouzov1973. Use it when you need the model to write or continue text. It is set up for mlx. The card lists the license as apache-2.0.
Reprise MLX 4-bit de Ministral-3-3B (Mistral AI, France — Apache-2.0), pour usage on-device texte dans Generex (assistant tactique iPad). Chargé via Mistral3TextModel (mlx-swift-lm ≥ 2.29.3) : le backbone texte 3B est…
All-time downloads
255
Public
Parameters
3.8B
2.8 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors2.7 GB · 99%
Parameter types
How the weights are stored.
U323.4B · 90%
Base models
Model card
From the Hugging Face model README
Ministral-3-3B-Instruct-2512 — MLX 4-bit (Generex)
Reprise MLX 4-bit de Ministral-3-3B (Mistral AI, France — Apache-2.0), pour usage
on-device texte dans Generex (assistant tactique iPad). Chargé via
Mistral3TextModel (mlx-swift-lm ≥ 2.29.3) : le backbone texte 3B est utilisé,
les poids de vision sont écartés au chargement (~1,8 Go en RAM → cible iPad 8 Go).
tokenizer_class corrigé TokenizersBackend → LlamaTokenizer pour compatibilité
swift-transformers (tokenizer BPE/ByteLevel, chat_template.jinja conservé).
- Task
- Text Generation
- Library
- mlx
- Type
- mistral3
- License
- apache-2.0
- Languages
- fr, en
- Created
- Jun 16, 2026
- Updated
- Jun 16, 2026