BGE Large EN v1.5
BAAI · BAAI/bge-large-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) · 1024d | 45.08 | independent run |
|---|
Reported by the publisher
- MTEB AmazonCounterfactualClassification (en) (en; test; accuracy)
- 75.851Classification
- MTEB AmazonPolarityClassification (test; accuracy)
- 92.417Classification
- MTEB AmazonReviewsClassification (en) (en; test; accuracy)
- 48.176Classification
- MTEB ArguAna (test; ndcg_at_10)
- 63.542Retrieval
- MTEB ArxivClusteringP2P (test; v_measure)
- 48.567Clustering
- MTEB ArxivClusteringS2S (test; v_measure)
- 43.195Clustering
- MTEB AskUbuntuDupQuestions (test; map)
- 64.466Reranking
- MTEB BIOSSES (test; cos_sim_spearman)
- 84.654STS
Every reported result
- MTEB Banking77Classification (test; accuracy)
- 87.789Classification
- MTEB BiorxivClusteringP2P (test; v_measure)
- 39.715Clustering
- MTEB BiorxivClusteringS2S (test; v_measure)
- 36.901Clustering
- MTEB CQADupstackAndroidRetrieval (test; ndcg_at_10)
- 50.218Retrieval
- MTEB CQADupstackEnglishRetrieval (test; ndcg_at_10)
- 49.039Retrieval
- MTEB CQADupstackGamingRetrieval (test; ndcg_at_10)
- 59.313Retrieval
- MTEB CQADupstackGisRetrieval (test; ndcg_at_10)
- 41.439Retrieval
- MTEB CQADupstackMathematicaRetrieval (test; ndcg_at_10)
- 33.158Retrieval
- MTEB CQADupstackPhysicsRetrieval (test; ndcg_at_10)
- 47.06Retrieval
- MTEB CQADupstackProgrammersRetrieval (test; ndcg_at_10)
- 42.659Retrieval
- MTEB CQADupstackRetrieval (test; ndcg_at_10)
- 42.229Retrieval
- MTEB CQADupstackStatsRetrieval (test; ndcg_at_10)
- 37.329Retrieval
- MTEB CQADupstackTexRetrieval (test; ndcg_at_10)
- 30.003Retrieval
- MTEB CQADupstackUnixRetrieval (test; ndcg_at_10)
- 42.1Retrieval
- MTEB CQADupstackWebmastersRetrieval (test; ndcg_at_10)
- 39.232Retrieval
- MTEB CQADupstackWordpressRetrieval (test; ndcg_at_10)
- 35.198Retrieval
- MTEB ClimateFEVER (test; ndcg_at_10)
- 36.567Retrieval
- MTEB DBPedia (test; ndcg_at_10)
- 44.106Retrieval
- MTEB EmotionClassification (test; accuracy)
- 51.520Classification
- MTEB FEVER (test; ndcg_at_10)
- 87.182Retrieval
- MTEB FiQA2018 (test; ndcg_at_10)
- 45.021Retrieval
- MTEB HotpotQA (test; ndcg_at_10)
- 74.101Retrieval
- MTEB ImdbClassification (test; accuracy)
- 92.846Classification
- MTEB MSMARCO (dev; ndcg_at_10)
- 42.493Retrieval
- MTEB MTOPDomainClassification (en) (en; test; accuracy)
- 94.592Classification
- MTEB MTOPIntentClassification (en) (en; test; accuracy)
- 79.494Classification
- MTEB MassiveIntentClassification (en) (en; test; accuracy)
- 77.562Classification
- MTEB MassiveScenarioClassification (en) (en; test; accuracy)
- 80.535Classification
- MTEB MedrxivClusteringP2P (test; v_measure)
- 32.561Clustering
- MTEB MedrxivClusteringS2S (test; v_measure)
- 31.470Clustering
- MTEB MindSmallReranking (test; map)
- 32.061Reranking
- MTEB NFCorpus (test; ndcg_at_10)
- 38.129Retrieval
- MTEB NQ (test; ndcg_at_10)
- 55.029Retrieval
- MTEB QuoraRetrieval (test; ndcg_at_10)
- 89.065Retrieval
- MTEB RedditClustering (test; v_measure)
- 57.242Clustering
- MTEB RedditClusteringP2P (test; v_measure)
- 64.662Clustering
- MTEB SCIDOCS (test; ndcg_at_10)
- 22.64Retrieval
- MTEB SICK-R (test; cos_sim_spearman)
- 81.683STS
- MTEB STS12 (test; cos_sim_spearman)
- 79.051STS
- MTEB STS13 (test; cos_sim_spearman)
- 86.374STS
- MTEB STS14 (test; cos_sim_spearman)
- 82.778STS
- MTEB STS15 (test; cos_sim_spearman)
- 88.031STS
- MTEB STS16 (test; cos_sim_spearman)
- 86.487STS
- MTEB STS17 (en-en) (en-en; test; cos_sim_spearman)
- 87.498STS
- MTEB STS22 (en) (en; test; cos_sim_spearman)
- 67.053STS
- MTEB STSBenchmark (test; cos_sim_spearman)
- 87.516STS
- MTEB SciDocsRR (test; map)
- 87.633Reranking
- MTEB SciFact (test; ndcg_at_10)
- 74.607Retrieval
- MTEB SprintDuplicateQuestions (test; cos_sim_ap)
- 96.726PairClassification
- MTEB SprintDuplicateQuestions (test; max_ap)
- 96.726PairClassification
- MTEB StackExchangeClustering (test; v_measure)
- 66.399Clustering
- MTEB StackExchangeClusteringP2P (test; v_measure)
- 34.964Clustering
- MTEB StackOverflowDupQuestions (test; map)
- 55.947Reranking
- MTEB SummEval (test; cos_sim_spearman)
- 31.606Summarization
- MTEB TRECCOVID (test; ndcg_at_10)
- 74.818Retrieval
- MTEB Touche2020 (test; ndcg_at_10)
- 24.806Retrieval
- MTEB ToxicConversationsClassification (test; accuracy)
- 70.914Classification
- MTEB TweetSentimentExtractionClassification (test; accuracy)
- 59.943Classification
- MTEB TwentyNewsgroupsClustering (test; v_measure)
- 51.266Clustering
- MTEB TwitterSemEval2015 (test; cos_sim_ap)
- 79.043PairClassification
- MTEB TwitterSemEval2015 (test; max_ap)
- 79.043PairClassification
- MTEB TwitterURLCorpus (test; cos_sim_ap)
- 85.600PairClassification
- MTEB TwitterURLCorpus (test; max_ap)
- 85.600PairClassification
- Sequence Length
- 512Evaluation · MTEB
- Average (56)
- 64.23Evaluation · MTEB
- Retrieval (15)
- 54.29Evaluation · MTEB
- Clustering (11)
- 46.08Evaluation · MTEB
- Pair Classification (3)
- 87.12Evaluation · MTEB
- Reranking (4)
- 60.03Evaluation · MTEB
- STS (10)
- 83.11Evaluation · MTEB
- Summarization (1)
- 31.61Evaluation · MTEB
- Classification (12)
- 75.97Evaluation · MTEB
Values as printed on the card, with the publisher's own settings (card section: Evaluation); column “BAAI/bge-large-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 Large EN v1.5 for retrieval and semantic search: it turns text into 1,024-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-large-en-v1.5')
model = AutoModel.from_pretrained('BAAI/bge-large-en-v1.5', revision="refs/pr/13")
model_ort = ORTModelForFeatureExtraction.from_pretrained('BAAI/bge-large-en-v1.5', revision="refs/pr/13",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-large-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-large-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 (335,142,400 parameters, read 2026-10-09 from the Hub API).
- Announced
- 2023-09-12
The card lists English only. PT-BR shows as confirmed solely through the Portuguese-only subsets of MTEB (Multilingual, v2), a small indicator computed from the public results dataset, not a publisher claim.
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.






























































