docs: announce multimodal sibling repo
Browse files
README.md
CHANGED
|
@@ -13,6 +13,13 @@ tags:
|
|
| 13 |
pipeline_tag: text-generation
|
| 14 |
---
|
| 15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
## Use it from Swift
|
| 17 |
|
| 18 |
<!-- swift-usage-begin -->
|
|
|
|
| 13 |
pipeline_tag: text-generation
|
| 14 |
---
|
| 15 |
|
| 16 |
+
> **🆕 Multimodal version available** — for text + image + video + audio,
|
| 17 |
+
> use [`mlboydaisuke/gemma-4-E4B-multimodal-coreml`](https://huggingface.co/mlboydaisuke/gemma-4-E4B-multimodal-coreml).
|
| 18 |
+
> Same Gemma 4 E4B decoder + ANE-targeted vision encoder + Conformer
|
| 19 |
+
> audio encoder. Validated 2026-05-03 on iPhone 17 Pro at 15.7 tok/s
|
| 20 |
+
> with all four input modalities working. This text-only repo stays
|
| 21 |
+
> available for users who don't need vision/audio.
|
| 22 |
+
|
| 23 |
## Use it from Swift
|
| 24 |
|
| 25 |
<!-- swift-usage-begin -->
|