Downloads · 30 days
18
5% of all-time downloads
ministral-3-3b-reasoning-mlx
sdkv2/ministral-3-3b-reasoning-mlx
ministral-3-3b-reasoning-mlx is a text generation model from sdkv2. 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.
MLX conversion of mistralai/Ministral-3-3B-Reasoning-2512 (Apache-2.0), used as the primary model in the multiheadtest copy-suppression circuit analysis project. Bit-exact replication verified against the stock model…
All-time downloads
369
Public
Parameters
3.4B
6.9 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors6.9 GB · 100%
At a glance
- Task
- Text Generation
- Library
- mlx
- License
- apache-2.0
- Model type
- mistral3
- Access
- Public
- Created
- Aug 7, 2026
- Updated
- Aug 7, 2026
- SHA
- fa7cb6e7
Try a prompt
Model card
From the Hugging Face model README
Ministral-3-3B-Reasoning (MLX, bf16)
MLX conversion of mistralai/Ministral-3-3B-Reasoning-2512 (Apache-2.0),
used as the primary model in the multiheadtest copy-suppression circuit
analysis project. Bit-exact replication verified against the stock model
(smoke check: max|dlogit| = 0.0 vs stock forward).
Notes
- 26 layers, 32 heads, GQA 4:1, head_dim 128, Llama-4-style RoPE scaling.
- Loaded with
fix_mistral_regex: Truetokenizer config in this project. - The q8 quantized variant is at
sdkv2/ministral-3-3b-reasoning-mlx-q8.
- Task
- Text Generation
- Library
- mlx
- Type
- mistral3
- License
- apache-2.0
- Languages
- en
- Created
- Aug 7, 2026
- Updated
- Aug 7, 2026