SAVRN
Search Contact SAVRN

SAVRN Model Hub · Comparisons

bge-base-en vs SapBERT-from-PubMedBERT-fulltext

Bge-base-en has 109M parameters and SapBERT-from-PubMedBERT-fulltext has 109M parameters; bge-base-en is released under MIT License and SapBERT-from-PubMedBERT-fulltext under Apache License 2.0; at 16-bit, bge-base-en needs about 0.3 GB (1x MI300X from $1.85 an hour) and SapBERT-from-PubMedBERT-fulltext about 0.3 GB (1x MI300X from $1.85 an hour).

Published metadata for 2 models, each read from its own repository.
Field bge-base-en
BAAI/bge-base-en
SapBERT-from-PubMedBERT-fulltext
cambridgeltl/SapBERT-from-PubMedBERT-fulltext
Publisher Beijing Academy of Artificial Intelligence Language Technology Lab @University of Cambridge
Task Feature extraction Feature extraction
Modality Text Text
Parameters, as reported 109M parameters 109M parameters
Architecture BertModel BertModel
Library transformers transformers
Context length 512 tokens 512 tokens
Repository size 1.3 GB 1.8 GB
Artifact formats safetensors, onnx, pytorch safetensors, pytorch, jax, tf
License mit apache-2.0
Access Open weights, no gate Open weights, no gate
Memory at 16-bit (weights and margin) 0.3 GB 0.3 GB
Cheapest GPUs at 16-bit, per hour 1x MI300X, $1.85 1x MI300X, $1.85
Memory at 4-bit (weights and margin) 0.1 GB 0.1 GB
Cheapest GPUs at 4-bit, per hour 1x MI300X, $1.85 1x MI300X, $1.85
Revision viewed b737bf5dcc6e 090663c3ae57
Downloads reported by the hub 2.5M 1.5M
Last observed 2026-09-18 2026-09-18

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.

bge-base-en

BenchmarkConditionsResultReported byRevisionDate
MTEB AmazonCounterfactualClassification (en) Configuration enTask ClassificationMetric accuracyComparison conditions not established 75.7313 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonCounterfactualClassification (en) Configuration enTask ClassificationMetric apComparison conditions not established 38.9728 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonCounterfactualClassification (en) Configuration enTask ClassificationMetric f1Comparison conditions not established 69.8174 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonPolarityClassification Configuration defaultTask ClassificationMetric accuracyComparison conditions not established 92.5652 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonPolarityClassification Configuration defaultTask ClassificationMetric apComparison conditions not established 88.8882 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonPolarityClassification Configuration defaultTask ClassificationMetric f1Comparison conditions not established 92.5482 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (en) Configuration enTask ClassificationMetric accuracyComparison conditions not established 46.91 BAAI
Publisher reported
Evaluated revision not stated
MTEB AmazonReviewsClassification (en) Configuration enTask ClassificationMetric f1Comparison conditions not established 46.2854 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric map_at_1Comparison conditions not established 38.834 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric map_at_10Comparison conditions not established 53.564 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric map_at_100Comparison conditions not established 54.23 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric map_at_1000Comparison conditions not established 54.235 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric map_at_3Comparison conditions not established 49.49 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric map_at_5Comparison conditions not established 51.784 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric mrr_at_1Comparison conditions not established 39.26 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric mrr_at_10Comparison conditions not established 53.744 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric mrr_at_100Comparison conditions not established 54.41 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric mrr_at_1000Comparison conditions not established 54.415 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric mrr_at_3Comparison conditions not established 49.656 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric mrr_at_5Comparison conditions not established 52.018 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_1Comparison conditions not established 38.834 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established 61.487 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_100Comparison conditions not established 64.303 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_1000Comparison conditions not established 64.408 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_3Comparison conditions not established 53.116 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric ndcg_at_5Comparison conditions not established 57.248 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric precision_at_1Comparison conditions not established 38.834 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric precision_at_10Comparison conditions not established 8.663 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric precision_at_100Comparison conditions not established 0.989 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric precision_at_1000Comparison conditions not established 0.1 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric precision_at_3Comparison conditions not established 21.219 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric precision_at_5Comparison conditions not established 14.737 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric recall_at_1Comparison conditions not established 38.834 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric recall_at_10Comparison conditions not established 86.629 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric recall_at_100Comparison conditions not established 98.862 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric recall_at_1000Comparison conditions not established 99.644 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric recall_at_3Comparison conditions not established 63.656 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArguAna Configuration defaultTask RetrievalMetric recall_at_5Comparison conditions not established 73.684 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArxivClusteringP2P Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 48.8848 BAAI
Publisher reported
Evaluated revision not stated
MTEB ArxivClusteringS2S Configuration defaultTask ClusteringMetric v_measureComparison conditions not established 42.8505 BAAI
Publisher reported
Evaluated revision not stated

SAVRN's Notes on bge-base-en

The publisher's own page steers you to bge-base-en-v1.5 as the replacement, so the first question is why this one. Usually the answer is a vector store already built on it; re-embedding is the switching cost. The hardware numbers barely register: 109 million parameters, 0.2 GB of weights at 16-bit, 0.3 GB to run. A 192 GB MI300X at $1.85 an hour is wasted on this alone; we run models this size as a sidecar beside a generation model. The 512-token context is the binding limit; anything longer gets chunked.

MIT asks for one thing, the copyright and permission notice, and gives commercial use, modification and redistribution in return. The 1.3 GB download carries three formats, safetensors, ONNX and PyTorch, in float32, so pick one. And check the dates: released August 5, 2023, last updated April 17, 2024, and with the publisher pointing elsewhere this is a dependency you maintain yourself.

SAVRN's Notes on SapBERT-from-PubMedBERT-fulltext

One vector per biomedical entity name is the entire output of SapBERT-from-PubMedBERT-fulltext, and that narrow job is why it draws 1.5 million downloads a month four years after its March 2, 2022 release. Cambridge's Language Technology Lab trained it on UMLS 2020AA, English only, on a PubMedBERT base, and the [CLS] embedding of the last layer is the feature you keep. At 109 million parameters it needs 0.3 GB at 16-bit and 0.1 GB at 8-bit or 4-bit, so an entity-linking service runs on a sliver of the $1.85-an-hour MI300X on the Index.

Apache 2.0 permits commercial use, modification and redistribution with the notices intact. Two checks before you build on it: the 512-token context means inputs are entity names, not passages, and the last update was June 14, 2023, with the method in arXiv:2010.11784. Weights ship in safetensors, pytorch, jax and tf, so the serving stack is your call.

Questions

Which is larger, bge-base-en or SapBERT-from-PubMedBERT-fulltext?

bge-base-en (109M parameters) is larger than SapBERT-from-PubMedBERT-fulltext (109M parameters), by the parameter counts their publishers report.

Which is cheaper to run, bge-base-en or SapBERT-from-PubMedBERT-fulltext?

At 4-bit, bge-base-en fits on 1x MI300X from $1.85 an hour and SapBERT-from-PubMedBERT-fulltext on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use bge-base-en commercially?

Yes. bge-base-en is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.

Can I use SapBERT-from-PubMedBERT-fulltext commercially?

Yes. SapBERT-from-PubMedBERT-fulltext 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