BGE Small EN v1.5
BAAI · BAAI/bge-small-en-v1.5
Weights license: MIT · code and components may use other terms
Quality
MTEB (Multilingual, v2)
Mean(task) · 0–100 · higher is better · this model outlined
- Color = publisher
Source: MTEB results dataset (independent runs, CC0) · publisher-reported where the model card is the only source · other benchmarks and versions are not mixed in
| MTEB (Multilingual, v2) | mean(task) · 384d | 43.76 | independent run |
|---|
Reported by the publisher
- MTEB AmazonCounterfactualClassification (en) (en; test; accuracy)
- 73.791Classification
- MTEB AmazonPolarityClassification (test; accuracy)
- 92.754Classification
- MTEB AmazonReviewsClassification (en) (en; test; accuracy)
- 46.986Classification
- MTEB ArguAna (test; ndcg_at_10)
- 59.550Retrieval
- MTEB ArxivClusteringP2P (test; v_measure)
- 47.403Clustering
- MTEB ArxivClusteringS2S (test; v_measure)
- 40.015Clustering
- MTEB AskUbuntuDupQuestions (test; map)
- 62.586Reranking
- MTEB BIOSSES (test; cos_sim_spearman)
- 83.755STS
Every reported result
- MTEB Banking77Classification (test; accuracy)
- 85.740Classification
- MTEB BiorxivClusteringP2P (test; v_measure)
- 38.467Clustering
- MTEB BiorxivClusteringS2S (test; v_measure)
- 34.719Clustering
- MTEB CQADupstackAndroidRetrieval (test; ndcg_at_10)
- 47.658Retrieval
- MTEB CQADupstackEnglishRetrieval (test; ndcg_at_10)
- 45.586Retrieval
- MTEB CQADupstackGamingRetrieval (test; ndcg_at_10)
- 56.379Retrieval
- MTEB CQADupstackGisRetrieval (test; ndcg_at_10)
- 38.378Retrieval
- MTEB CQADupstackMathematicaRetrieval (test; ndcg_at_10)
- 29.701Retrieval
- MTEB CQADupstackPhysicsRetrieval (test; ndcg_at_10)
- 43.922Retrieval
- MTEB CQADupstackProgrammersRetrieval (test; ndcg_at_10)
- 38.231Retrieval
- MTEB CQADupstackRetrieval (test; ndcg_at_10)
- 39.052Retrieval
- MTEB CQADupstackStatsRetrieval (test; ndcg_at_10)
- 34.482Retrieval
- MTEB CQADupstackTexRetrieval (test; ndcg_at_10)
- 27.386Retrieval
- MTEB CQADupstackUnixRetrieval (test; ndcg_at_10)
- 38.769Retrieval
- MTEB CQADupstackWebmastersRetrieval (test; ndcg_at_10)
- 37.462Retrieval
- MTEB CQADupstackWordpressRetrieval (test; ndcg_at_10)
- 30.673Retrieval
- MTEB ClimateFEVER (test; ndcg_at_10)
- 31.836Retrieval
- MTEB DBPedia (test; ndcg_at_10)
- 40.029Retrieval
- MTEB EmotionClassification (test; accuracy)
- 47.845Classification
- MTEB FEVER (test; ndcg_at_10)
- 86.643Retrieval
- MTEB FiQA2018 (test; ndcg_at_10)
- 40.343Retrieval
- MTEB HotpotQA (test; ndcg_at_10)
- 69.935Retrieval
- MTEB ImdbClassification (test; accuracy)
- 90.606Classification
- MTEB MSMARCO (dev; ndcg_at_10)
- 40.833Retrieval
- MTEB MTOPDomainClassification (en) (en; test; accuracy)
- 93.361Classification
- MTEB MTOPIntentClassification (en) (en; test; accuracy)
- 74.747Classification
- MTEB MassiveIntentClassification (en) (en; test; accuracy)
- 74.808Classification
- MTEB MassiveScenarioClassification (en) (en; test; accuracy)
- 78.699Classification
- MTEB MedrxivClusteringP2P (test; v_measure)
- 33.059Clustering
- MTEB MedrxivClusteringS2S (test; v_measure)
- 30.856Clustering
- MTEB MindSmallReranking (test; map)
- 31.290Reranking
- MTEB NFCorpus (test; ndcg_at_10)
- 34.305Retrieval
- MTEB NQ (test; ndcg_at_10)
- 50.178Retrieval
- MTEB QuoraRetrieval (test; ndcg_at_10)
- 88.78Retrieval
- MTEB RedditClustering (test; v_measure)
- 52.318Clustering
- MTEB RedditClusteringP2P (test; v_measure)
- 60.640Clustering
- MTEB SCIDOCS (test; ndcg_at_10)
- 20.522Retrieval
- MTEB SICK-R (test; cos_sim_spearman)
- 79.410STS
- MTEB STS12 (test; cos_sim_spearman)
- 77.441STS
- MTEB STS13 (test; cos_sim_spearman)
- 82.980STS
- MTEB STS14 (test; cos_sim_spearman)
- 81.836STS
- MTEB STS15 (test; cos_sim_spearman)
- 87.264STS
- MTEB STS16 (test; cos_sim_spearman)
- 84.933STS
- MTEB STS17 (en-en) (en-en; test; cos_sim_spearman)
- 87.151STS
- MTEB STS22 (en) (en; test; cos_sim_spearman)
- 65.300STS
- MTEB STSBenchmark (test; cos_sim_spearman)
- 85.865STS
- MTEB SciDocsRR (test; map)
- 85.938Reranking
- MTEB SciFact (test; ndcg_at_10)
- 71.275Retrieval
- MTEB SprintDuplicateQuestions (test; cos_sim_ap)
- 96.670PairClassification
- MTEB SprintDuplicateQuestions (test; max_ap)
- 96.678PairClassification
- MTEB StackExchangeClustering (test; v_measure)
- 60.780Clustering
- MTEB StackExchangeClusteringP2P (test; v_measure)
- 35.273Clustering
- MTEB StackOverflowDupQuestions (test; map)
- 53.641Reranking
- MTEB SummEval (test; cos_sim_spearman)
- 30.121Summarization
- MTEB TRECCOVID (test; ndcg_at_10)
- 75.899Retrieval
- MTEB Touche2020 (test; ndcg_at_10)
- 26.037Retrieval
- MTEB ToxicConversationsClassification (test; accuracy)
- 69.886Classification
- MTEB TweetSentimentExtractionClassification (test; accuracy)
- 60.507Classification
- MTEB TwentyNewsgroupsClustering (test; v_measure)
- 48.521Clustering
- MTEB TwitterSemEval2015 (test; cos_sim_ap)
- 73.239PairClassification
- MTEB TwitterSemEval2015 (test; max_ap)
- 73.328PairClassification
- MTEB TwitterURLCorpus (test; cos_sim_ap)
- 84.839PairClassification
- MTEB TwitterURLCorpus (test; max_ap)
- 84.984PairClassification
- Sequence Length
- 512Evaluation · MTEB
- Average (56)
- 62.17Evaluation · MTEB
- Retrieval (15)
- 51.68Evaluation · MTEB
- Clustering (11)
- 43.82Evaluation · MTEB
- Pair Classification (3)
- 84.92Evaluation · MTEB
- Reranking (4)
- 58.36Evaluation · MTEB
- STS (10)
- 81.59Evaluation · MTEB
- Summarization (1)
- 30.12Evaluation · MTEB
- Classification (12)
- 74.14Evaluation · MTEB
Values as printed on the card, with the publisher's own settings (card section: Evaluation); column “BAAI/bge-small-en-v1.5”. Not comparable across publishers and not used in any ranking. Model card
Quality × memory
No comparable quality × memory pair
Embedding memory estimates cluster at a few GB, so a quality × memory chart would not separate these models; MTEB is ranked above instead.
Consider BGE Small EN v1.5 for retrieval and semantic search: it turns text into 384-dimension vectors.
The publisher states no memory figure; the hardware list uses a ~3 GB planning estimate, a ceiling and not a measurement. See the prompt format under How to run. PT-BR is confirmed by a source.
Runs on
Fast memory holds it 119
The planning estimate fits the memory the model can use quickly: graphics card VRAM, or unified memory within the macOS GPU share.
Show all 119
Hardware profiles grouped by the embedding planning estimate (8-bit weights, 20% allowance, 2 GB reserve). Not a measured fit.
Measured speed
No published measurement for this checkpoint
Throughput (texts per second) is shown only when a source publishes it with hardware, build and setup. It is never estimated.
How to run
The planning estimate uses 8-bit weights, 20% working allowance and a 2 GB reserve; it is a ceiling, not a measurement.
- Runs in
- Sentence Transformerssource
- Query/document prompts
- Optional query instruction 'Represent this sentence for searching relevant passages: ' for short-query retrieval; documents take no prefix. For v1.5 the card says retrieval works without the instruction with only a slight degradation.
- Languages · PT-BR
- EnglishPT-BR confirmed by a source
Documented ways to run it
Copied as printed in the linked source (read 2026-10-18); not run by this site.
from optimum.onnxruntime import ORTModelForFeatureExtraction # type: ignore
import torch
from transformers import AutoModel, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-small-en-v1.5')
model = AutoModel.from_pretrained('BAAI/bge-small-en-v1.5')
model_ort = ORTModelForFeatureExtraction.from_pretrained('BAAI/bge-small-en-v1.5', file_name="onnx/model.onnx")
# Sentences we want sentence embeddings for
sentences = ["样例数据-1", "样例数据-2"]
# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
# for s2p(short query to long passage) retrieval task, add an instruction to query (not add instruction for passages)
# encoded_input = tokenizer([instruction + q for q in queries], padding=True, truncation=True, return_tensors='pt')
model_output_ort = model_ort(**encoded_input)
# Compute token embeddings
with torch.no_grad():
model_output = model(**encoded_input)
# model_output and model_output_ort are identical
import asyncio
from infinity_emb import AsyncEmbeddingEngine, EngineArgs
sentences = ["Embed this is sentence via Infinity.", "Paris is in France."]
engine = AsyncEmbeddingEngine.from_args(
EngineArgs(model_name_or_path = "BAAI/bge-small-en-v1.5", device="cpu", engine="optimum" # or engine="torch"
))
async def main():
async with engine:
embeddings, usage = await engine.embed(sentences=sentences)
asyncio.run(main())Technical details
What the source and the catalog store for this checkpoint. Missing fields are left out, never zero.
- Checkpoint / source
- BAAI/bge-small-en-v1.5
- Task
- embeddings for search (feature extraction)
- Library in source
- Sentence Transformers
- Weight license
- MIT
- Commercial use
- Allowed MIT license text (SPDX) · read 2026-10-18license text
- Parameter basis
- Hub safetensors total (33,360,512 parameters, read 2026-10-09 from the Hub API).
- Announced
- 2023-09-12
The MTEB results metadata for this model lists 512 dimensions while the card states 384, so no MTEB score or Portuguese indicator is attributed to it until the two agree.
Sources
Identity and published specifications come from the checkpoint page. Benchmark scores belong to the named release and do not measure speed on your computer.
- Publisher model card
- Hugging Face model page / card · 2026-10-09
- Hugging Face API (parameters, licence, library) · 2026-10-09
- MTEB results dataset (CC0), scores computed by scripts/ingest-mteb-results.mjs · 2026-10-25
Curated facts checked: 2026-10-09. Each value cites its source (the publisher's card or the MTEB results dataset); estimates are labelled as estimates, and a value no source states stays empty. A repository commit is not a model release date.






























































