Jina AI

Introducing jina-embeddings-v5-text. Our fifth-generation multilingual embedding models in two efficient sizes: a 677M small and 239M nano model — with task-specific LoRA adapters, Matryoshka dimensions, 32K context, and GGUF/MLX quantization for edge deployment, setting new benchmarks across MMTEB, MTEB English, and retrieval tasks.

After releasing Jina VLM, we tore apart 70+ VLM and noticed something interesting. Language models have scaled to hundreds of billions of parameters, but 𝐯𝐢𝐬𝐢𝐨𝐧 𝐞𝐧𝐜𝐨𝐝𝐞𝐫𝐬? Still the same handful. And most people still only know CLIP. Our 30-page survey covers the landscape of 𝐯𝐢𝐬𝐢𝐨𝐧 𝐞𝐧𝐜𝐨𝐝𝐞𝐫𝐬: three training paradigms, dynamic resolution handling, multi-encoder fusion, and whether encoder-free architectures actually work. Useful if you're picking vision encoders or just want to understand how today's VLM got here.

Releasing jina-VLM: our new 2B vision language model achieves SOTA on multilingual visual question answering and document understanding among open 2B-scale VLMs. Available now via API & HuggingFace!

Our 3rd BoF on Embeddings, Rerankers, Small LMs for Better Search at #EMNLP2025, Suzhou - after Singapore and Miami. 100+ attendees, 8 speakers featuring Andrianos Michail, Lucas Moeller, Ziyang Zeng, Hyukkyu Kang, Marc Briner, Siyue ZHANG, and Saba Sturua.

“Jina AI’s team and technology bring cutting-edge models into the Elastic ecosystem, making our platform even more powerful for context engineering. Together, we are expanding what developers and enterprises can achieve with search-powered AI, while staying true to our commitment to openness and accessibility.” — Ashutosh Kulkarni, Elastic CEO. Today, we are excited to announce that we have joined forces with Jina AI, a pioneer in open source multimodal and multilingual embeddings, reranker, and small language models. This acquisition deepens Elastic’s capabilities in vector search, RAG, and context engineering to power agentic AI. Read more: https://go.es.io/4nBwD5L

Heard you like GGUFs and MLX. Our newly released listwise reranker, jina-reranker-v3, is now available in dynamic quantized GGUFs and MLX.

New 0.6B-parameter listwise reranker that considers the query and all candidate documents in a single context window. https://lnkd.in/duXy27p9

We brought multimodal embeddings to llama.cpp and GGUF, and uncovered a few surprising issues along the way. https://lnkd.in/e7EU_k7r

Today we're releasing jina-code-embeddings, a new suite of code embedding models in two sizes—0.5B and 1.5B parameters—along with 1~4bit GGUF quantizations for both. Built on latest code generation LLMs, these models achieve SOTA retrieval performance despite their compact size. They support over 15 programming languages and 5 tasks: nl2code, code2code, code2nl, code2completions and qa.

Code generation LLMs → code embeddings: 0.5B/1.5B models achieve SOTA performance across 25 code retrieval benchmarks. https://lnkd.in/g2T54-Yp