Company

jina-embeddings-v2-base-en

Jina AI

jina-embeddings-v2-base-en is a 137-million-parameter English text embedding model released by Jina AI in November 2023, distinguished as the first open-source model to support 8K (8192 token) input length at a time when only OpenAI's proprietary text-embedding-ada-002 offered comparable context capacity . The model outputs 768-dimensional vectors and was designed for high-precision, large-scale tasks, with a smaller sibling (jina-embeddings-v2-small-en at 33 million parameters and 512-dimensional output) available for lightweight deployments . On MTEB benchmarks, it reportedly outperformed OpenAI's rival model on text classification, retrieval, reranking, and summarization tasks while offering developers lower compute and storage costs .

Jina AI itself was founded by CEO Xiao Han (肖涵) and seed-funded by Yunqi Capital, which led its seed round and remained an investor through its acquisition by Elastic in October 2025 . The Berlin-based startup positioned the model as a building block for retrieval-augmented generation (RAG) systems, where long-context embedding helps compress and query documents without losing semantic granularity across lengthy texts like legal filings or medical literature .

AI-generated — may contain errors, please verify.

jina-embeddings-v2-base-enCompany
Jina AI
渲染中…
Mentioned in 33 articles

Coverage