Professional model card

README.md

@@ -1,24 +1,24 @@
 ---
 license: apache-2.0
 base_model: Qwen/Qwen3-4B
+library_name: basert
+pipeline_tag: text-generation
 tags:
   - basert
-  - base-format
   - apple-silicon
   - quantized
 ---
 
-# Qwen3-4B (BaseRT `.base`)
+# Qwen3-4B
 
-[BaseRT](https://github.com/basecompute/baseRT) `.base` conversions of
-[`Qwen/Qwen3-4B`](https://huggingface.co/Qwen/Qwen3-4B) for the BaseRT Apple-Silicon (Metal)
-inference engine. Converted **directly from the bf16 source** (no
-quant-from-quant) to avoid compounding quantization error.
+[BaseRT](https://github.com/basecompute/baseRT) `.base` builds of [`Qwen/Qwen3-4B`](https://huggingface.co/Qwen/Qwen3-4B) for fast local inference on Apple Silicon (Metal).
 
-| File | Quant | Size |
-|------|-------|------|
-| `Qwen3-4B-Q4.base` | `base_q4` | 2.1G |
-| `Qwen3-4B-Q8.base` | `base_q8` | 3.9G |
+## Files
+
+| File | Precision | Size |
+|------|-----------|------|
+| `Qwen3-4B-Q4.base` | 4-bit | 2.3 GB |
+| `Qwen3-4B-Q8.base` | 8-bit | 4.2 GB |
 
 ## Usage
 
@@ -28,4 +28,4 @@ basert pull basecompute/Qwen3-4B
 basert chat basecompute/Qwen3-4B
 ```
 
-Context: 32k · License inherited from the base model.
+Released under the apache-2.0 license, inherited from the base model.