bge-micro-v2 by Taylor: Open-Weight Model

Runs On

What it takes to serve bge-micro-v2 (17M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

Precision	Cheapest setup	Per hour	Also fits
16-bit	1x MI300X (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x MI325X $2.00
8-bit	1x MI300X (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x MI325X $2.00
4-bit	1x MI300X (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on bge-micro-v2

Three transformer layers, 17 million parameters, and 156 MB of float32 weights: the publisher distilled bge-micro-v2 in two steps from BAAI/bge-small-en-v1.5, and what remains maps sentences and paragraphs into a 384-dimensional space. At 16-bit our memory line rounds to 0.0 GB. We price the cheapest slot at one MI300X with 192 GB for $1.85 an hour because that is the lowest line we carry, not because the model needs it; it rides on a card you already pay for.

MIT permits commercial use, modification and redistribution with the notices kept, so shipping it inside a product is paperwork only. Two checks before committing: the 512-token context caps each passage you embed, and the file lists no relation records, so the lineage to bge-small-en-v1.5 rests on the publisher's description alone. Downloads run 1,058,661 monthly, the last update was 2024-06-06, and no host price sits on the Index yet.

SAVRN Research, 2026-09-18

Model Card

By Taylor, published under mit, revision 3edf6d7de0fa.

This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search.

Distilled in a 2-step training process (bge-micro was step 1) from BAAI/bge-small-en-v1.5.

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('{MODEL_NAME}')
embeddings = model.encode(sentences)
print(embeddings)

Usage (HuggingFace Transformers)

Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.

Read the full model card (273 words)

Configuration

Architecture: BertModel
Context length (tokens): 512
Layers: 3
Hidden size: 384
Feed-forward size: 1,536
Attention heads: 12
Vocabulary size: 30,522
Stored precision: float32
Model type: bert

Identity and Version

Repository: TaylorAI/bge-micro-v2
Publisher: Taylor
Task: Sentence similarity
Modality: Text
Library: sentence-transformers
Parameters: 17M parameters
Languages: Not stated by the source
Revision: 3edf6d7de0faa426b09780416fe61009f26ae589
First published: 2023-10-11
Last updated: 2024-06-06

Files and Weights

17 files, 157.0 MB in total. The weights are 4 files totalling 156.0 MB in bin, onnx, safetensors.

Weights4 files · 156.0 MB

Configuration7 files · 1.7 KB

Tokenizer3 files · 944.7 KB

Documentation2 files · 66.6 KB

Repository1 file · 1.5 KB

Every file

File	Type	Size	SHA-256
model.safetensors	Weights	34.8 MB	792472b64c3c
onnx/model.onnx	Weights	69.0 MB	9f705befe60d
onnx/model_quantized.onnx	Weights	17.4 MB	ed65e36025aa
pytorch_model.bin	Weights	34.8 MB	f154c669357b
1_Pooling/config.json	Configuration	270 B	—
added_tokens.json	Configuration	82 B	—
config.json	Configuration	745 B	—
config_sentence_transformers.json	Configuration	123 B	—
modules.json	Configuration	229 B	—
sentence_bert_config.json	Configuration	53 B	—
special_tokens_map.json	Configuration	228 B	—
LICENSE	Documentation	1.1 KB	—
README.md	Documentation	65.5 KB	—
.gitattributes	Repository	1.5 KB	—
tokenizer.json	Tokenizer	711.7 KB	—
tokenizer_config.json	Tokenizer	1.6 KB	—
vocab.txt	Tokenizer	231.5 KB	—

License and Download

License: mit
Access: Open weights, no gate
Download size: 156.0 MB

Download from Taylor

Released by Taylor through its official repository on Hugging Face. Read the license.

Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

Benchmark	Conditions	Result	Reported by	Revision	Date
MTEB AmazonCounterfactualClassification (en)	Configuration enTask ClassificationMetric accuracyComparison conditions not established	67.7612	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonCounterfactualClassification (en)	Configuration enTask ClassificationMetric apComparison conditions not established	29.6378	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonCounterfactualClassification (en)	Configuration enTask ClassificationMetric f1Comparison conditions not established	61.3118	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonPolarityClassification	Configuration defaultTask ClassificationMetric accuracyComparison conditions not established	79.7547	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonPolarityClassification	Configuration defaultTask ClassificationMetric apComparison conditions not established	74.214	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonPolarityClassification	Configuration defaultTask ClassificationMetric f1Comparison conditions not established	79.6532	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonReviewsClassification (en)	Configuration enTask ClassificationMetric accuracyComparison conditions not established	37.452	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB AmazonReviewsClassification (en)	Configuration enTask ClassificationMetric f1Comparison conditions not established	37.0245	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric map_at_1Comparison conditions not established	31.152	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric map_at_10Comparison conditions not established	46.702	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric map_at_100Comparison conditions not established	47.563	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric map_at_1000Comparison conditions not established	47.567	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric map_at_3Comparison conditions not established	42.058	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric map_at_5Comparison conditions not established	44.608	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric mrr_at_1Comparison conditions not established	32.006	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric mrr_at_10Comparison conditions not established	47.064	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric mrr_at_100Comparison conditions not established	47.91	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric mrr_at_1000Comparison conditions not established	47.915	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric mrr_at_3Comparison conditions not established	42.283	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric mrr_at_5Comparison conditions not established	44.968	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric ndcg_at_1Comparison conditions not established	31.152	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric ndcg_at_10Comparison conditions not established	55.308	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric ndcg_at_100Comparison conditions not established	58.965	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric ndcg_at_1000Comparison conditions not established	59.067	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric ndcg_at_3Comparison conditions not established	45.698	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric ndcg_at_5Comparison conditions not established	50.296	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric precision_at_1Comparison conditions not established	31.152	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric precision_at_10Comparison conditions not established	8.279	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric precision_at_100Comparison conditions not established	0.987	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric precision_at_1000Comparison conditions not established	0.1	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric precision_at_3Comparison conditions not established	18.753	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric precision_at_5Comparison conditions not established	13.485	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric recall_at_1Comparison conditions not established	31.152	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric recall_at_10Comparison conditions not established	82.788	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric recall_at_100Comparison conditions not established	98.72	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric recall_at_1000Comparison conditions not established	99.502	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric recall_at_3Comparison conditions not established	56.259	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArguAna	Configuration defaultTask RetrievalMetric recall_at_5Comparison conditions not established	67.425	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArxivClusteringP2P	Configuration defaultTask ClusteringMetric v_measureComparison conditions not established	44.5269	TaylorAI Publisher reported	Evaluated revision not stated	—
MTEB ArxivClusteringS2S	Configuration defaultTask ClusteringMetric v_measureComparison conditions not established	33.2457	TaylorAI Publisher reported	Evaluated revision not stated	—

Memory Requirements

Precision	Weights in memory
As published	156.0 MB
16-bit	0.0 GB
8-bit	0.0 GB
4-bit	0.0 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About bge-micro-v2

How much GPU memory does bge-micro-v2 need?

About 0 GB at 16-bit and 0 GB at 4-bit: the weights (17M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run bge-micro-v2 on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use bge-micro-v2 commercially?

Yes. bge-micro-v2 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.

What is bge-micro-v2's context length?

512 tokens, from the maximum position embeddings in its published configuration.

Similar Models

Model · Sentence similarity

paraphrase-MiniLM-L3-v2

Sentence Transformers

This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search. Using this model becomes easy when you have sentence-transformers installed: Then you can use the model like this: Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings. This model was trained by sentence-transformers. If you find this model helpful, feel free to cite our publication Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

Open weights apache-2.0 17M parameters 512 tokens sentence-transformers

View model

Model · Sentence similarity

all-MiniLM-L6-v2

Sentence Transformers

This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search. Using this model becomes easy when you have sentence-transformers installed: Then you can use the model like this: Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings. The project aims to train sentence embedding models on very large sentence level datasets using a self-supervised contrastive learning objective. We used the pretrained…

Open weights apache-2.0 23M parameters 512 tokens sentence-transformers

View model

Model · Sentence similarity

paraphrase-MiniLM-L6-v2

Sentence Transformers

This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search. Using this model becomes easy when you have sentence-transformers installed: Then you can use the model like this: Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings. This model was trained by sentence-transformers. If you find this model helpful, feel free to cite our publication Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

Open weights apache-2.0 23M parameters 512 tokens sentence-transformers

View model

Model · Sentence similarity

multi-qa-MiniLM-L6-cos-v1

Sentence Transformers

This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and was designed for semantic search. It has been trained on 215M (question, answer) pairs from diverse sources. For an introduction to semantic search, have a look at: SBERT.net - Semantic Search Using this model becomes easy when you have sentence-transformers installed: Then you can use the model like this: Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the correct pooling-operation on-top of the contextualized word embeddings. Similarly to the PyTorch example above, to use the model…

Open weights 23M parameters 512 tokens sentence-transformers

View model

Model · Sentence similarity

paraphrase-albert-small-v2

Sentence Transformers

This is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search. Using this model becomes easy when you have sentence-transformers installed: Then you can use the model like this: Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings. This model was trained by sentence-transformers. If you find this model helpful, feel free to cite our publication Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

Open weights apache-2.0 12M parameters 512 tokens sentence-transformers

View model

Model · Sentence similarity

rubert-tiny2

David Dale

This is an updated version of cointegrated/rubert-tiny: a small Russian BERT-based encoder with high-quality sentence embeddings. This post in Russian gives more details. The differences from the previous version include: - sentence embeddings approximate LaBSE closer than before; - meaningful segment embeddings (tuned on the NLI task) - the model is focused only on Russian. The model should be used as is to produce sentence embeddings (e.g. for KNN classification of short texts) or fine-tuned for a downstream task. Sentence embeddings can be produced as follows: Alternatively, you can use the model with sentencetransformers: For those who want to run the inference with VLLM, there is a…

Open weights mit 29M parameters 2,048 tokens sentence-transformers

View model