janihal/embeddinggemma-300m-8bit-hi-mlx

Add 'Compared to other embedding models' section (Qwen3-Embedding 0.6B/4B, Qwen3-VL-Embedding-2B)

MLX mixed-precision 8-bit (blocks int8 g32, embedding/Dense fp16) of google/embeddinggemma-300m + model card with FP32/GGUF benchmarks

initial commit