Upload README.md with huggingface_hub

README.md

@@ -19,32 +19,55 @@ extra_gated_description: If you want to learn more about how we process your per
 tags:
 - mistral-common
 - mlx
+- ministral
+- ministral-3
+- vision-language
+- multimodal
+- quantized
+- edge
+- 4-bit
+- base-model
 pipeline_tag: image-text-to-text
 base_model: mistralai/Ministral-3-14B-Base-2512
 ---
 
 # mlx-community/Ministral-3-14B-Base-2512-4bit
 
-The largest model in the Ministral 3 family, **Ministral 3 14B Base 2512** is a vision-language model: a text backbone paired
+This is, **Ministral 3 14B Base 2512** is a vision-language model: a text backbone paired
 with a vision encoder, supporting image understanding alongside text. This is
 the **base pre-trained** checkpoint — not instruction- or chat-tuned. For
 chat/instruction-following use cases, use the
 [Instruct variant](https://huggingface.co/mlx-community/Ministral-3-14B-Instruct-2512-4bit)
 instead; this base checkpoint is intended for custom post-training/fine-tuning.
 
+> **Community note.** Structural check confirms the vision tower and
+> multimodal projector were carried over intact (not dropped, which is a real
+> failure mode for text-only conversion tools on vision-language models).
+> Functional check confirms both text-only and image+text generation produce
+> coherent output. Converted and verified by a single maintainer running
+> local MLX tooling -- not independently reviewed by anyone else; please open
+> a discussion if you hit anything unexpected.
+
 This is an MLX conversion of [`mistralai/Ministral-3-14B-Base-2512`](https://huggingface.co/mistralai/Ministral-3-14B-Base-2512),
 converted with [mlx-vlm](https://github.com/Blaizzy/mlx-vlm). Refer to the
 [original model card](https://huggingface.co/mistralai/Ministral-3-14B-Base-2512) for the full
 description, capabilities, and license terms.
 
-## Quantization notes
+## Heads up
+
+- **Base model, not instruct-tuned** — expect raw completion behavior, not
+  chat-following. Don't expect it to follow instructions well.
+- **Vision retained at full precision** — only the language backbone is
+  quantized; the vision tower and multimodal projector are untouched bf16,
+  per mlx-vlm's standard policy of not quantizing multimodal modules.
+- **Output size on disk: 9.47GB**
+
+## Provenance
 
-- Language model layers: 4-bit, group_size=64, affine mode
-- Vision tower and multimodal projector are kept at **full precision** — only
-  the language backbone is quantized, per mlx-vlm's standard policy of not
-  quantizing multimodal modules
+- Source: [`mistralai/Ministral-3-14B-Base-2512`](https://huggingface.co/mistralai/Ministral-3-14B-Base-2512) (BF16)
+- Language model layers: **4-bit** affine quantization, group_size=64
+- Vision tower + multimodal projector: kept at full precision (not quantized)
 - Blended average: **5.416 bits per weight** across all parameters
-- Output size on disk: **9.47GB**
 
 ## Ministral 3 family