Jina AI

jina-embeddings-v5-omni vv5-omni

jina-embeddings-v5-omni introduces frozen-encoder model composition for text-geometry-preserving multimodal embeddings. It covers text, images, audio, and video in a unified embedding space, sharing the exact same text space as jina-embeddings-v5-text. Available in small and nano sizes.

jina-embeddings-v5-text v5

Jina AI's 'jina-embeddings-v5-text' is the fifth generation of their multilingual embedding models, designed to deliver state-of-the-art performance in tasks such as retrieval, text matching, clustering, and classification. Released on February 18, 2026, this model offers two variants: 'jina-embeddings-v5-text-small' with 677 million parameters and 'jina-embeddings-v5-text-nano' with 239 million parameters, both supporting context lengths up to 32,768 tokens.