The story so far
Weights, every dated article, and repo commits on one rail.
17 repo commits since the weights landed
Teal is the weights (2023-05-19), grey ticks are repo commits.
e5-small-v2
33.4M parameters, text in, embeddings out, 512 tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #43 in Embeddings · score 90.5 method v1 Sources 3 Downloads 30d 1.1M · #106 of 619 tracked
Do people use it?1.1Mdownloads, 30 days#106 of 619 we track · top 17%+2 likes, downloads +19% this week136 likesHugging Face2026-10-24
Where to run it
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | HF Inference | — | — | — | — | Oct 25 |
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 1 API provider; 15 quantized builds for local use.
- As a service1HF Inference
- On own hardware15quantized builds
- Built on4946 fine-tunes · 3 adapters · 0 merges
- In apps97Hugging Face Spaces
Top Spaces running it: radames/Candle-BERT-Semantic-Similarity-Wasm ♥ 17 · amryassin/embedding-bench ♥ 2 · riccorl/relik-entity-linking ♥ 1 · Zeitstaub/AI-Patents_searched_by_AI ♥ 1
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at BF16
About 0.1 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.
#27 of 64 on MTEB so far
61.32. 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
What people built on it
64 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
64 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.
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 | 48.46 nDCG@10 x100 (mean of 10 tasks) | 10 tasks, split=test | leaderboard | 2026-10-24 |
| measured MTEB · Embed text for search | 61.32 mean(task) x100 over 41 tasks | 41 tasks, split=test | leaderboard | 2026-10-24 |
| measured MTEB Reranking · Rerank search results | 44.1 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 produces an embedding size of 384.[1]
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.
Facts
21 fields
Listed in Embeddings
Facts
- e5-small-v2 has 33.4M 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: 33,360,000 · 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-small-v2: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 0.1 GB · Edge (<4 GB) |
|---|---|
| FP8 (1 byte/param) | 0.0 GB · Edge (<4 GB) |
| INT4 (0.5 bytes/param) | 0.0 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
All variants
40 variants
Lineage
finetune: 46quantized: 15adapter: 3
Top variants (40 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| alana89/TabSTAR | finetune | — | 47.3M | S | cc-by-4.0 | alana89 | 31 | 16.8M |
| Xenova/e5-small-v2 | quantized | — | — | — | — | Xenova | 4 | 195.2k |
| OpenMed/OpenMed-PII-ClinicalE5-Small-33M-v1 | finetune | — | 33.3M | S | apache-2.0 | OpenMed | 0 | 79.5k |
| ChristianAzinn/e5-small-v2-gguf | quantized | gguf | — | — | mit | ChristianAzinn | 1 | 6.6k |
| mradermacher/e5-small-v2-i1-GGUF | quantized | GGUF | — | — | mit | mradermacher | 0 | 4.3k |
| Bombek1/e5-small-v2-litert | finetune | — | — | — | mit | Bombek1 | 0 | 3.4k |
| mradermacher/e5-small-v2-GGUF | quantized | GGUF | — | — | mit | mradermacher | 0 | 2.4k |
| ggml-org/e5-small-v2-Q8_0-GGUF | quantized | Q8_0 | — | — | mit | ggml-org | 1 | 1.5k |
| Krelle/e5-small-v2-imo-pairs | finetune | — | 33.4M | S | — | Krelle | 0 | 913 |
| clips/e5-small-v2-t2t | finetune | — | 33.2M | S | mit | clips | 0 | 621 |
| JumareKhaled/e5-small-v2-medichat-finetuned | finetune | — | 33.4M | S | — | JumareKhaled | 1 | 360 |
| takara-ai/e5-small-v2-static-distilled | finetune | — | 7.56M | S | mit | takara-ai | 0 | 401 |
| dgwon/e5-small-v2.nfcorpus.meta-llama-Llama-3.1-8B-Instruct | finetune | — | 33.4M | S | — | dgwon | 0 | 369 |
| jangedoo/e5-small-v2-pruned | finetune | — | 30.2M | S | — | jangedoo | 0 | 326 |
| dgwon/e5-small-v2.nfcorpus.meta-llama-Llama-3.1-8B-Instruct.confilt | finetune | — | 33.4M | S | — | dgwon | 0 | 317 |
| juanmcristobal/e5-small-v2-mitre | finetune | — | 33.4M | S | — | juanmcristobal | 0 | 286 |
| mbudisic/e5-small-v2-ft-pstuts | finetune | — | 33.4M | S | — | mbudisic | 0 | 277 |
| OpenMed/OpenMed-PII-French-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 244 |
| benjac8/biobite-retriever | finetune | — | 33.4M | S | mit | benjac8 | 0 | 224 |
| OpenMed/OpenMed-PII-Spanish-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 201 |
| OpenMed/OpenMed-PII-German-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 188 |
| thegenerativegeneration/stay_or_go_conversation_classifier_xs | finetune | — | 33.4M | S | — | thegenerativegeneration | 0 | 178 |
| mbudisic/e5-small-v2-20250519015557 | finetune | — | 33.4M | S | — | mbudisic | 0 | 173 |
| thomasht86/e5-small-v2-Q4_K_M-GGUF | quantized | GGUF | — | — | mit | thomasht86 | 0 | 171 |
| OpenMed/OpenMed-PII-Portuguese-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 163 |
| dgwon/e5-small-v2.nfcorpus.meta-llama-Llama-3.2-3B-Instruct.top10pct | finetune | — | 33.4M | S | — | dgwon | 0 | 150 |
| SudhaBhattacharjee/RewardModel_deberta-v3-large | finetune | — | 33.4M | S | mit | SudhaBhattacharjee | 1 | 99 |
| OpenMed/OpenMed-PII-Italian-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 142 |
| OpenMed/OpenMed-PII-Dutch-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 139 |
| pineee/e5-small-v2-hivetrace-finetune-CE-infoNCE | finetune | — | 33.4M | S | mit | pineee | 2 | 35 |
| dgwon/e5-small-v2.nfcorpus.meta-llama-Llama-3.1-8B-Instruct.top10pct | finetune | — | 33.4M | S | — | dgwon | 0 | 125 |
| rockon1095/e5-small-v2-Q8_0-GGUF | quantized | Q8_0 | — | — | mit | rockon1095 | 0 | 104 |
| dgwon/e5-small-v2.nfcorpus.meta-llama-Llama-3.2-3B-Instruct | finetune | — | 33.4M | S | — | dgwon | 0 | 103 |
| mbotsu/e5-small-v2-lora | adapter | — | — | — | — | mbotsu | 0 | 102 |
| OpenMed/OpenMed-PII-Hindi-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 94 |
| rockon1095/e5-small-v2-Q4_0-GGUF | quantized | Q4_0 | — | — | mit | rockon1095 | 0 | 93 |
| dgwon/e5-small-v2.nfcorpus.meta-llama-Llama-3.2-3B-Instruct.confilt | finetune | — | 33.4M | S | — | dgwon | 0 | 92 |
| abdulrahman-nuzha/intfloat-e5-small-v2-english-fp16-allagree | finetune | -fp16 | 33.4M | S | mit | abdulrahman-nuzha | 0 | 91 |
| OpenMed/OpenMed-PII-Turkish-ClinicalE5-Small-33M-v1 | finetune | — | 33.2M | S | apache-2.0 | OpenMed | 0 | 90 |
| jrc2139/e5-small-v2-ONNX | quantized | — | — | — | — | jrc2139 | 0 | 88 |
Repository changelog
17 commits
Repository changelog 17 commits
| Date | Commit | SHA |
|---|---|---|
| 2025-02-17 | Add exported openvino model 'openvino_model_qint8_quantized.xml' (#15) | ffb93f3 |
| 2025-02-17 | Add exported openvino model 'openvino_model.xml' (#14) | 4bc52f9 |
| 2025-02-17 | Add exported onnx model 'model_O4.onnx' (#13) | d908e34 |
| 2025-02-17 | Add exported onnx model 'model_qint8_avx512_vnni.onnx' (#12) | f17a0f7 |
| 2023-08-16 | Make the tokenizer config match that of the large and base versions (#11) | dca8b1a |
| 2023-08-07 | Update README.md | 93da57d |
| 2023-07-27 | add sentence similarity tag | b092564 |
| 2023-07-20 | Add TF weights (#3) | 4a074f4 |
| 2023-07-20 | Adding ONNX file of this model (#1) | 47ba724 |
| 2023-07-20 | Adding `safetensors` variant of this model (#7) | 5c28755 |
| 2023-07-20 | Update README.md | cfb7694 |
| 2023-06-18 | add sentence transformers configs (#4) | 7f5bd82 |
| 2023-05-22 | Update README.md | c2a786b |
| 2023-05-19 | Update README.md | 6cf6a8c |
| 2023-05-19 | update README.md | e32cfbb |
| 2023-05-19 | upload model weights | d7c6f2a |
| 2023-05-19 | initial commit | b584ff7 |
All providers
1 provider · 1 endpoint
Where to run 1 providers · 1 endpoints
Hosted endpoints observed for e5-small-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.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| HF Inference hf | intfloat/e5-small-v2 | — | — | — | — | — | — | — | live | 2026-10-25 |
Articles and coverage
2 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.
Analysis2
webai.show842 words
E5-small-v2 is a 33M-parameter embedder that outputs 384-dimensional vectors and requires distinct prefixes for input. It uses a query: prefix for searches and a passage: prefix for documents to optimize retrieval performance. The model runs on-device via WebAssembly with q8 quantization at ~34 MB under an MIT license.written, reviewed, cited
- E5 mean-pools token vectors and L2-normalizes them so every sentence becomes a unit-length arrow.
- The model loads MiniLM, BGE, and GTE on demand to show how four embedders disagree on similarity.
- Inference runs in a Web Worker so the matrix, projection, and comparison never jank the page.
savrn.com2,650 words
e5-small-v2 is an open-weight sentence-transformers model by Liang Wang designed for semantic search and clustering tasks. It has 33M parameters and a 512-token context, mapping text to 384-dimensional vectors under the MIT License. At 16-bit precision it needs about 0.1 GB of GPU memory, fitting on 1x MI300X from $1.85 an hour.written, reviewed, cited
- The stored precision is float32, so the 16-bit figure assumes conversion on load.
- The publisher's usage prefixes inputs with query: and passage:, and an index built without those prefixes is not the same index.
- The model was first published on 2023-05-19 and last updated on 2025-02-17.
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).
Similar models
3 models
Similar models
- paraphrase-MiniLM-L6-v2 · 22.7M same size band (S), same task (sentence-similarity), more providers, permissive license
- multi-qa-MiniLM-L6-cos-v1 · 22.7M same size band (S), same task (sentence-similarity)
- all-distilroberta-v1 · 82.1M same size band (S), same task (sentence-similarity), permissive license
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:36 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:04 UTC · Last verified 2026-10-25
Follow e5-small-v2
Watch this model
- /feeds/models/intfloat--e5-small-v2.xml(RSS 2.0) · every change to e5-small-v2
- /feeds/families/intfloat--e5-small-v2.xml(RSS 2.0) · every change across the e5-small-v2 family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.