jina-embeddings-v2-base-en is a 137M-parameter English text embedding model with an 8,192-token context window that natively handles long documents without truncation.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
No API provider serves it, so its downloads are people running it themselves: 10 quantized builds and 59 Spaces.
Third-party measurements only; author-reported numbers stay in Reference, marked as such.
#18 of 33 on MTEB Reranking so far
46.39. Each tick is a model on this site with a measured MTEB Reranking row; author-reported numbers stay in Reference. MAP x100 (mean of 2 tasks). 2 tasks, split=test
Everything else, folded: the full tables, every source.
Benchmarks
3 rows
Benchmarks 3 rows
measured third-party leaderboard row (3)author-reported number from the model card or eval filesNo confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
jina-embeddings-v2-base-en is an English monolingual embedding model with 137M parameters.[1][2] The model uses a BERT architecture and supports an 8192 sequence length.[3][4]
Generated 2026-10-07 15:04 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Use-case evidence
2 use cases
What it is good at 2 use cases
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
jina-embeddings-v2-base-en is a 137M-parameter English text embedding model with an 8,192-token context window that natively handles long documents without truncation. It produces 768-dimensional vectors from a 307MB footprint and outperformed OpenAI's text-embedding-ada-002 on classification, reranking, retrieval, and summarization tasks. While jina-embeddings-v5-text-small (677M params, 32K context) outperforms it for most production workloads in 2026, this model remains relevant for lightweight English-only pipelines.written, reviewed, cited
aws.amazon.com2024-06-09about the jina embeddings v2 familyHN 1 pts · 0 comments1,732 words
Jina Embeddings v2 supports an 8,192-tokens context length for processing long documents like legal text or product documentation. The jina-embeddings-v2-base-de model has a size of 322 MB with a performance score of 60.1% for information retrieval tasks.generated, cited
plushcap.comabout the jina embeddings v2 family213 words
Jina AI released jina-embeddings-v2-base-en, an open-source text embedding model matching OpenAI's ada-002 performance in benchmarks and context window length. The model is currently only available for English text but can be deployed via Jina AI's model library or Truss. It supports applications such as search, recommendation, clustering, and retrieval-augmented generation.written, reviewed, cited
Jina AI released jina-embeddings-v2-base-en in October 2024 as a specialized English embedding model supporting sequences up to 8192 tokens. Trained with 137 million parameters, the model generates 768-dimensional vectors and is available for free under the Apache 2.0 license. Engineers can integrate it via PyMilvus or the SentenceTransformer library to generate embeddings for semantic search.written, reviewed, cited
github.com2024-03-24authorityabout the jina embeddings v2 family548 words
Jina-embeddings-v2-base-en is an English monolingual embedding model with 8192 sequence length built on BERT architecture (JinaBERT). It supports a symmetric bidirectional variant of ALiBi for extended sequence length and was pretrained on the C4 dataset. The model underwent further training on Jina AI's collection of 400M+ sentence pairs and hard negatives from diverse domains.written, reviewed, cited
ritvik19.medium.com2024-12-03about the jina embeddings v2 family
Papers
2 papers
Papers 2
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
