SAVRN
Search Contact SAVRN

SAVRN Model Hub · Comparisons

gte-multilingual-base vs paraphrase-multilingual-mpnet-base-v2

Gte-multilingual-base has 305M parameters and paraphrase-multilingual-mpnet-base-v2 has 278M parameters; both are released under Apache License 2.0; at 16-bit, gte-multilingual-base needs about 0.7 GB (1x MI300X from $1.85 an hour) and paraphrase-multilingual-mpnet-base-v2 about 0.7 GB (1x MI300X from $1.85 an hour).

Published metadata for 2 models, each read from its own repository.
Field gte-multilingual-base
Alibaba-NLP/gte-multilingual-base
paraphrase-multilingual-mpnet-base-v2
sentence-transformers/paraphrase-multilingual-mpnet-base-v2
Publisher Alibaba-NLP Sentence Transformers
Task Sentence similarity Sentence similarity
Modality Text Text
Parameters, as reported 305M parameters 278M parameters
Architecture NewModel XLMRobertaModel
Library sentence-transformers sentence-transformers
Context length 8,192 tokens 514 tokens
Repository size 629.6 MB 10.9 GB
Artifact formats safetensors safetensors, onnx, openvino, pytorch, tf
License apache-2.0 apache-2.0
Access Open weights, no gate Open weights, no gate
Memory at 16-bit (weights and margin) 0.7 GB 0.7 GB
Cheapest GPUs at 16-bit, per hour 1x MI300X, $1.85 1x MI300X, $1.85
Memory at 4-bit (weights and margin) 0.2 GB 0.2 GB
Cheapest GPUs at 4-bit, per hour 1x MI300X, $1.85 1x MI300X, $1.85
Revision viewed 9bbca17d9273 4328cf26390c
Downloads reported by the hub 1.4M 9.8M
Last observed 2026-09-19 2026-09-19

An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.

Other Reported Results

These results are listed for each model on its own, because the conditions needed to compare them are not stated or do not match. Two results that leave a condition blank are not assumed to share it.

gte-multilingual-base

BenchmarkConditionsResultReported byRevisionDate
MTEB 8TagsClustering Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 33.6668 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AFQMC Configuration defaultTask STSMetric cos_sim_spearmanComparison conditions not established 43.5476 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB ATEC Configuration defaultTask STSMetric cos_sim_spearmanComparison conditions not established 48.9119 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AllegroReviews Configuration defaultTask ClassificationMetric accuracyComparison conditions not established 41.6899 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AlloProfClusteringP2P Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 54.2024 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AlloProfClusteringS2S Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 44.3408 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AlloprofReranking Configuration defaultTask RerankingMetric mapComparison conditions not established 64.915 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AlloprofRetrieval Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 53.638 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonCounterfactualClassification (en) Configuration enTask ClassificationMetric accuracyComparison conditions not established 75.9552 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonPolarityClassification Configuration defaultTask ClassificationMetric accuracyComparison conditions not established 80.7176 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (de) Configuration deTask ClassificationMetric accuracyComparison conditions not established 40.108 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (en) Configuration enTask ClassificationMetric accuracyComparison conditions not established 43.642 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (es) Configuration esTask ClassificationMetric accuracyComparison conditions not established 40.17 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (fr) Configuration frTask ClassificationMetric accuracyComparison conditions not established 39.568 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (ja) Configuration jaTask ClassificationMetric accuracyComparison conditions not established 35.75 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (zh) Configuration zhTask ClassificationMetric accuracyComparison conditions not established 33.342 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 58.231 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB ArguAna-PL Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 53.166 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB ArxivClusteringP2P Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 46.019 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB ArxivClusteringS2S Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 41.0663 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB AskUbuntuDupQuestions Configuration defaultTask RerankingMetric mapComparison conditions not established 61.8751 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BIOSSES Configuration defaultTask STSMetric cos_sim_spearmanComparison conditions not established 81.2145 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BQ Configuration defaultTask STSMetric cos_sim_spearmanComparison conditions not established 51.7159 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BSARDRetrieval Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 26.115 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BUCC (de-en) Configuration de-enTask BitextMiningMetric f1Comparison conditions not established 98.6169 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BUCC (fr-en) Configuration fr-enTask BitextMiningMetric f1Comparison conditions not established 97.896 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BUCC (ru-en) Configuration ru-enTask BitextMiningMetric f1Comparison conditions not established 97.1239 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BUCC (zh-en) Configuration zh-enTask BitextMiningMetric f1Comparison conditions not established 98.1569 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB Banking77Classification Configuration defaultTask ClassificationMetric accuracyComparison conditions not established 85.3604 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BiorxivClusteringP2P Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 37.5904 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB BiorxivClusteringS2S Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 34.2147 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CBD Configuration defaultTask ClassificationMetric accuracyComparison conditions not established 62.52 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CDSC-E Configuration defaultTask PairClassificationMetric cos_sim_apComparison conditions not established 74.9013 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CDSC-R Configuration defaultTask STSMetric cos_sim_spearmanComparison conditions not established 90.3073 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CLSClusteringP2P Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 37.9485 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CLSClusteringS2S Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 38.1196 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CMedQAv1 Configuration defaultTask RerankingMetric mapComparison conditions not established 86.1095 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CMedQAv2 Configuration defaultTask RerankingMetric mapComparison conditions not established 87.2804 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CQADupstackAndroidRetrieval Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 47.099 Alibaba-NLP
Publisher reported
Evaluated revision not stated
MTEB CQADupstackEnglishRetrieval Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 45.973 Alibaba-NLP
Publisher reported
Evaluated revision not stated

paraphrase-multilingual-mpnet-base-v2

BenchmarkConditionsResultReported byRevisionDate
mteb/arguana Task ArguAnaMetric ArguAnaSetup Obtained using MTEB v1.12.75Comparison conditions not established 48.908 Obtained using MTEB v1.12.75
Reported by a third party
Evaluated revision not stated 2026-03-05
mteb/arguana Task ArguAna_default_testMetric ArguAna_default_testSetup Obtained using MTEB v1.12.75Comparison conditions not established 48.908 Obtained using MTEB v1.12.75
Reported by a third party
Evaluated revision not stated 2026-03-05

SAVRN's Notes on gte-multilingual-base

Embedding the documents behind a retrieval system is where we would put this one. At 16-bit the weights take 0.6 GB and working memory 0.7 GB, so hardware is not the decision. The cheapest setup in our data, one MI300X with 192 GB at $1.85 an hour on-demand, would sit well under one percent occupied. The 8,192-token window covers long passages, and a 250,048-entry vocabulary says it was built for many languages.

Apache 2.0 permits commercial use, modification and redistribution, so an index built from it is yours to sell access to, as long as the notices travel with any copy. Before committing, read the paper trail: arXiv:2407.19669 describes the model, and the file also points at arXiv:2402.03216, the BGE-M3 paper, and arXiv:2104.08663, the BEIR benchmark. No host prices appear in the SAVRN Index for this model; running it yourself is the price.

SAVRN's Notes on paraphrase-multilingual-mpnet-base-v2

This is one of the most used multilingual embedding models: about 9.8 million downloads a month. It maps sentences in more than 50 languages into one shared vector space, so a query in one language can find a match in another.

At 278M parameters it runs on a CPU. Apache 2.0 allows commercial use. It is a dependable default for multilingual semantic search and clustering; newer embedding models in our catalog score higher on benchmarks, so test them if retrieval quality is your bottleneck.

Questions

Which is larger, gte-multilingual-base or paraphrase-multilingual-mpnet-base-v2?

gte-multilingual-base (305M parameters) is larger than paraphrase-multilingual-mpnet-base-v2 (278M parameters), by the parameter counts their publishers report.

Which is cheaper to run, gte-multilingual-base or paraphrase-multilingual-mpnet-base-v2?

At 4-bit, gte-multilingual-base fits on 1x MI300X from $1.85 an hour and paraphrase-multilingual-mpnet-base-v2 on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use gte-multilingual-base commercially?

Yes. gte-multilingual-base is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Can I use paraphrase-multilingual-mpnet-base-v2 commercially?

Yes. paraphrase-multilingual-mpnet-base-v2 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Related Comparisons