Research & releases
Keep building.
A reading list of Jina research and releases, with short independent Chatsax summaries and direct links to the source.
24 reading notes · Source checked September 7, 2026
jina-reranker-v3.5: Faster Listwise Reranking with Hybrid Attention and Self-Distillation
Explains the v3.5 architecture, distillation approach, and its role in document reranking.
jina-embeddings-v5-omni: Embeddings for Text, Image, Audio and Video
Introduces the v5-omni family and its shared representations across four input modalities.
Bootstrapping Audio Embeddings from Multimodal LLMs
Examines how multimodal generative models can be adapted into audio encoders for retrieval.
Identifying Embedding Models from Raw Numerical Values
Investigates whether numerical vector patterns can reveal which embedding model produced them.
jina-embeddings-v5-text: New SOTA Small Multilingual Embeddings
Presents two v5-text model sizes and deployment options for multilingual retrieval.
Jina-VLM: Small Multilingual Vision Language Model
Introduces a vision-language model designed for multilingual visual question answering.