Update README.md

README.md

@@ -16,15 +16,6 @@ If you intend to use this model outside of React Native ExecuTorch, make sure yo
 
 The **MLX** variant requires a physical Apple Silicon device (it does not run on the iOS simulator). The **XNNPACK** variant runs everywhere.
 
-## Variant Matrix
-
-| Delegate | Precision | File                                                       | Size    | Notes                                                                                                                                          |
-|----------|-----------|------------------------------------------------------------|---------|------------------------------------------------------------------------------------------------------------------------------------------------|
-| XNNPACK  | 8da4w     | `xnnpack/lfm_2_5_embedding_350m_xnnpack_8da4w.pte`         | 431 MB  | Int8 dynamic activation + Int4 weight (torchao), group_size=32, fp32 compute. Works on Android / iOS / generic CPU.                            |
-| MLX      | int4      | `mlx/lfm_2_5_embedding_350m_mlx_int4.pte`                  | 287 MB  | Int4 weight (group_size=64) with bf16 compute. Apple GPU; smallest variant. Requires a physical Apple Silicon device.                          |
-
-Both variants reproduce the upstream fp32 embedding with cosine ≈ 0.97 on a held-out set. Pick the variant that matches your platform; the MLX variant is iOS-only.
-
 ## Repository Structure
 
 - `xnnpack/` — `.pte` file partitioned for the XNNPACK delegate.