e5-base-v2

intfloat MIT out 3 years ago

109M parameters, text in, embeddings out, 512 tokens of context. 2 providers serve it from $0.005 per million input tokens.

Verified 2026-10-25 Rank #23 in Embeddings · score 96.3 method v1 Sources 3 Downloads 30d 964.5k · #115 of 619 tracked

Changed this week: 2 new quantized builds on the Hub (16 in all).2026-10-24

Can I run it?0.1GB at INT4Estimatefits a 16 GB cardmethod

What does it cost?$0.005/1M incheapest of 1DeepInfra2026-10-24

Can I use it?MITPermissivelicence file2026-10-24

Does it work?62.67on MTEB#22 of 64 measuredleaderboard2026-10-24

Do people use it?964.5kdownloads, 30 days#115 of 619 we track · top 19%+4 likes, downloads +9% this week171 likesHugging Face2026-10-24

Worth reading

Picked by rule from 4 tracked pieces: the lab's own post, the most shared, a field report, the latest.

4 best of 4 tracked

The 1 pick above, plus these 3. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleE5-base-v2 by intfloat | The GTM Directorythegtmdirectory.com · undated
ArticleThe intfloat E5 family guide: e5-base-v2, e5-large ... - Superlinkedsuperlinked.com · undated
Articlee5-base-v2 - Pinecone Docsdocs.pinecone.io · undated

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

#	Cheapest 2	In /1M	Out /1M	Context	Quant	Seen
1	DeepInfra	$0.005	—	512	—	Oct 24
2	HF Inference	—	—	—	—	Oct 25

All 2 providers →

Where it's used

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.

Served by 2 API providers; 16 quantized builds for local use; a big base for builders (81 fine-tunes, adapters and merges).

As a service2DeepInfra, HF Inference
On own hardware16quantized builds
Built on8181 fine-tunes · 0 adapters · 0 merges
In apps100+Hugging Face Spaces

Top Spaces running it: nickmuchi/semantic-search-with-retrieve-and-rerank ♥ 71 · radames/Candle-BERT-Semantic-Similarity-Wasm ♥ 17 · Tarka-AIR/Tarka-Embedding ♥ 4 · ipepe/nomic-embeddings ♥ 2

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 16 GB card at INT4

About 0.1 GB at INT4, 0.1 GB at FP8, 0.3 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

How it scores

Third-party measurements only; author-reported numbers stay in Reference, marked as such.

#22 of 64 on MTEB so far

62.67. Each tick is a model on this site with a measured MTEB row; author-reported numbers stay in Reference. Mean(task) x100 over 41 tasks. 41 tasks, split=test

Source · leaderboard · 2026-10-24

The story so far

Weights, every dated article, and repo commits on one rail.

14 repo commits since the weights landed

Teal is the weights (2023-05-19), grey ticks are repo commits.

What people built on it

97 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

97 derivatives on the Hub, most of them fine-tunes

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF ChristianAzinn/e5-base-v2-gguf for llama.cpp, Ollama, LM Studio 7.4k downloads

All variants →

Reference

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.

Benchmark	Value	Setting / notes	Reported by	Date
measured MTEB · Embed text for search	49.67 nDCG@10 x100 (mean of 10 tasks)	10 tasks, split=test	leaderboard	2026-10-24
measured MTEB · Embed text for search	62.67 mean(task) x100 over 41 tasks	41 tasks, split=test	leaderboard	2026-10-24
measured MTEB Reranking · Rerank search results	45.13 MAP x100 (mean of 2 tasks)	2 tasks, split=test	leaderboard	2026-10-24

What the card says

1 cited sentence

Summary generated, cited

The architecture consists of 12 layers and an embedding size of 768.^[1]

“This model has 12 layers and the embedding size is 768.”

Generated 2026-10-07 15:02 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.

Facts

21 fields

Listed in Embeddings

Facts

e5-base-v2 has 109M parameters. (safetensors metadata, observed 2026-10-24)
Context length is 512 tokens (512). (config.json max_position_embeddings, observed 2026-10-24)
License: MIT (friction: none). Permissive. (license file, observed 2026-10-24)
Created 2023-05-19, last modified 2025-02-17. (HF API, observed 2026-10-24)

Pipeline tag	sentence-similarity
Library	sentence-transformers
Architectures	BertModel
Model type	bert
Dtype	float32
Params by dtype	F32: 109,482,240 · I64: 512
Size band	S
MoE	no
Context length	512 (config.json max_position_embeddings)
Languages	en
Papers	arXiv:2212.03533, arXiv:2104.08663, arXiv:2210.07316
License	MIT (mit)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · BertModel · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · BertModel · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for e5-base-v2: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	0.3 GB · Edge (<4 GB)
FP8 (1 byte/param)	0.1 GB · Edge (<4 GB)
INT4 (0.5 bytes/param)	0.1 GB · Edge (<4 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Original model Family: intfloat/e5-v2 97 derivatives on the Hub

finetune: 81quantized: 16

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
OpenMed/OpenMed-PII-Italian-ClinicalE5-Base-109M-v1	finetune	—	109M	S	apache-2.0	OpenMed	0	115.2k
Xenova/e5-base-v2	quantized	—	—	—	—	Xenova	0	52.6k
mjwong/e5-base-v2-mnli-anli	finetune	—	109M	S	cc-by-nc-4.0	mjwong	0	14k
OpenMed/OpenMed-PII-ClinicalE5-Base-109M-v1	finetune	—	109M	S	apache-2.0	OpenMed	0	8.9k
ChristianAzinn/e5-base-v2-gguf	quantized	gguf	—	—	mit	ChristianAzinn	2	7.4k
mradermacher/e5-base-v2-i1-GGUF	quantized	GGUF	—	—	mit	mradermacher	0	4k
sapienzanlp/zebra-retriever-e5-base-v2	finetune	—	109M	S	cc	sapienzanlp	4	2.1k
mradermacher/e5-base-v2-GGUF	quantized	GGUF	—	—	mit	mradermacher	0	2.3k
Babelscape/wsl-retriever-e5-base-v2	finetune	—	—	—	cc-by-nc-sa-4.0	Babelscape	3	2k
clips/e5-base-v2-t2t	finetune	—	109M	S	mit	clips	0	1.6k
johnnas12/e5-galaxy-finetuned	finetune	—	109M	S	—	johnnas12	0	1.3k
rodekruis/sml-ukr-message-classifier-2	finetune	—	109M	S	—	rodekruis	1	1.2k
liuwenhan/Agentic-R_e5	finetune	—	109M	S	mit	liuwenhan	5	617
sieckenwingz/e5-resume-matcher-mnrl-v5	finetune	—	109M	S	—	sieckenwingz	0	855
wjunwei/ecommerce_text_embedding_retrieval_v2	finetune	—	109M	S	—	wjunwei	0	792
raghunath1/Aptivra-Base-110M	quantized	—	109M	S	mit	raghunath1	2	531
sapienzanlp/rbr-retriever-gkb-omcs-atomic	finetune	—	33.4M	S	cc	sapienzanlp	3	389
wjunwei/ecommerce_text_embedding_retrieval	finetune	—	109M	S	—	wjunwei	0	504
ElijahDevPH/e5-resume-matcher-realnav3	finetune	—	109M	S	—	ElijahDevPH	2	362
sivarohit2002/qwen06b_bi-e5-ft-weighted	finetune	—	109M	S	—	sivarohit2002	0	452
sapienzanlp/zebra-kb	finetune	—	—	—	cc	sapienzanlp	3	265
mangopy/ToolRet-trained-e5-base-v2	finetune	—	109M	S	—	mangopy	0	387
jamie8johnson/e5-base-v2-code-search	quantized	—	—	—	apache-2.0	jamie8johnson	0	357
rishhh/verisci-scifact-e5-retriever	finetune	—	109M	S	—	rishhh	0	357
drbinna/e5-nba-finetuned	finetune	—	109M	S	—	drbinna	0	353
shaswatamitra/falcon-yara-bi-e5-base-v2	finetune	—	109M	S	cc-by-nc-nd-4.0	shaswatamitra	0	353
vaktibabat/heart-e5	finetune	—	109M	S	—	vaktibabat	0	345
sieckenwingz/e5-resume-matcher-mnrl-freeze4-v5	finetune	—	109M	S	—	sieckenwingz	0	341
ElijahDevPH/e5-resume-matcher-mnrl	finetune	—	109M	S	—	ElijahDevPH	0	340
kimquynh/finetuned_sbert_4	finetune	—	109M	S	—	kimquynh	0	336
sowthri-industrial-ai/blending-embedder-e5	finetune	—	109M	S	—	sowthri-industrial-ai	0	314
sieckenwingz/e5-resume-matcher-mnrl-v3	finetune	—	109M	S	—	sieckenwingz	0	311
dpshade22/hf-e5-bible-100	finetune	—	109M	S	—	dpshade22	0	296
sieckenwingz/e5-resume-matcher-epoch1	finetune	—	109M	S	—	sieckenwingz	0	293
OpenMed/OpenMed-PII-French-ClinicalE5-Base-109M-v1	finetune	—	109M	S	apache-2.0	OpenMed	0	291
dpshade22/hf-e5-bible-150	finetune	—	109M	S	—	dpshade22	0	289
dpshade22/e5-bible-200	finetune	—	109M	S	—	dpshade22	0	286
sapienzanlp/zebra-kb-wg-train	finetune	—	—	—	cc	sapienzanlp	3	131
dpshade22/e5-base-bible	finetune	—	109M	S	—	dpshade22	0	274
nskwal/rayumi-5epochs	finetune	—	109M	S	—	nskwal	0	273

Repository changelog

14 commits

changed in the last 7 daysRank band moves: 2026-10-24Other: 2026-10-24all changes

Repository changelog 14 commits

Date	Commit	SHA
2025-02-17	Add exported openvino model 'openvino_model_qint8_quantized.xml' (#15)	f52bf8e
2025-02-17	Add exported openvino model 'openvino_model.xml' (#14)	e5f9ffa
2025-02-17	Add exported onnx model 'model_O4.onnx' (#13)	6d89b15
2025-02-17	Add exported onnx model 'model_qint8_avx512_vnni.onnx' (#12)	a079a01
2023-09-27	Adding ONNX file of this model (#11)	1c644c9
2023-08-07	Update README.md	cf5b12e
2023-07-27	add sentence similarity tag	6ad1cd1
2023-07-20	Adding `safetensors` variant of this model (#7)	5073c80
2023-07-20	Update README.md	2007ee8
2023-06-18	sentence_transformers (#5)	a3a4bc1
2023-05-22	Update README.md	ca2fa2a
2023-05-19	Update README.md	f9f7c80
2023-05-19	upload model weights	0f4dc66
2023-05-19	initial commit	350baba

All providers

2 providers · 2 endpoints

Where to run 2 providers · 2 endpoints

Hosted endpoints observed for e5-base-v2, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.0050/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
DeepInfra deepinfra	intfloat/e5-base-v2	512	—	$0.0050	—	—	—	—	live	2026-10-24
HF Inference hf	intfloat/e5-base-v2	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

4 pieces

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.

Analysis4

E5-base-v2 by intfloat | The GTM Directory

thegtmdirectory.com1,507 words

intfloat E5-base-v2 is an open-weight English embedding model with a 512-token context window and an MIT license. The model features approximately 110M parameters and 768-dimensional output, trained via weakly-supervised contrastive methods for asymmetric and symmetric retrieval tasks. Input costs $0.005 per 1M tokens on OpenRouter, and inputs must carry a 'query: ' or 'passage: ' prefix to avoid measurable accuracy loss.written, reviewed, cited

The model requires document chunking upstream because its 512-token hard limit prevents processing longer inputs without truncation.
Omitting the mandatory 'query: ' or 'passage: ' prefixes causes measurable performance degradation in retrieval accuracy.
The large variant offers higher retrieval accuracy with 24 layers and approximately 340M parameters at the cost of increased compute.

The intfloat E5 family guide: e5-base-v2, e5-large ... - Superlinked

superlinked.com727 words

The intfloat E5 family provides MIT-licensed embedding models including e5-base-v2 and multilingual-e5-large, all of which truncate inputs at 512 tokens. The multilingual-e5-large-instruct variant costs $0.0076 per million tokens and utilizes an XLM-RoBERTa-large backbone covering roughly 100 languages. Serving speeds reach 53,200 tokens/sec for base-v2 on a single L4 GPU, but omitting required role prefixes causes severe performance degradation.written, reviewed, cited

The e5-base-v2 parameter count is not stated on its model card, though 109M is the standard BERT-base figure.
The instruct variant replaces the prefix scheme by using a one-sentence task instruction format for queries.

e5-base-v2 - Pinecone Docs

docs.pinecone.io339 words

The e5-base-v2 model is an open-source embedding tool designed for good performance with minimal hardware requirements. It operates with 768 dimensions using a cosine metric and requires transformers==4.35.0 for installation. Users must prefix passages with 'passage: ' and queries with 'query: ' to achieve optimal results.written, reviewed, cited

e5-base-v2 — text embeddings score and price · Undominated.ai

undominated.ai225 words

Microsoft's e5-base-v2 is an open weights embedding model available from DeepInfra and OpenRouter. The model costs $0.0050 per 1M input tokens and covers 33 of the 41 tasks in the MTEB(eng, v2) benchmark. Consequently, it has no mean score and is not ranked on the benchmark.written, reviewed, cited

Papers

3 papers

Papers 3

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).

MTEB: Massive Text Embedding Benchmark cited in card

Niklas Muennighoff, Nouamane Tazi, Loïc Magne +1 · 2022-10-13 · PDF · HF paper page (9 upvotes)
BEIR: A Heterogenous Benchmark for Zero-shot Evaluation of Information Retrieval Models cited in card

Nandan Thakur, Nils Reimers, Andreas Rücklé +2 · 2021-04-17 · PDF · HF paper page (5 upvotes)
Text Embeddings by Weakly-Supervised Contrastive Pre-training cited in card

Liang Wang, Nan Yang, Xiaolong Huang +5 · 2022-12-07 · PDF · HF paper page (2 upvotes)

Similar models

3 models

Similar models

all-mpnet-base-v2 · 109M
same size band (S), same task (sentence-similarity), permissive license
paraphrase-multilingual-MiniLM-L12-v2 · 118M
same size band (S), same task (sentence-similarity), fewer providers, permissive license
multi-qa-mpnet-base-dot-v1 · 109M
same size band (S), same task (sentence-similarity)

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:35 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:02 UTC · Last verified 2026-10-25

Follow e5-base-v2

Watch this model

/feeds/models/intfloat--e5-base-v2.xml(RSS 2.0) · every change to e5-base-v2
/feeds/families/intfloat--e5-v2.xml(RSS 2.0) · every change across the e5-v2 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.