jinaai/jina-embeddings-v5-omni-nano-retrieval-mlx

readme: add frontier plot + EIS section

add omni_frontier.png: omni model frontier plot from paper

readme: add logo, ArXiv/Blog links, sibling-size link, broaden tags

Add Multimodal-parity-vs-torch section (fp32 torch ref, MLX with fp32 language_model)

Update Precision row + Precision note: bf16 vision/audio, fp32 language_model at load time

Upcast language_model bf16->fp32 at load time + tighten Conv1d transpose guard

Harmonize README structure with v5-omni GGUF cards (no umbrella)

docs: add Matryoshka, Video quickstart, and Selective modality sections

docs: drop stale mm-model pooling comparison from README

model.py: fp32 attention + rotary + LayerNorm (tighten bf16 drift) (#1)

MLX conversion v1: vision + text (no audio); re-keyed from torch base

initial commit