Bge Multilingual Gemma2

Jul 2024
Bge Multilingual Gemma2
Compare in
on
$0.11per 1M tokens
86MTEB English v2
Jul 2024released
3,584Dimensions
8K tokensMax input

Rankingconfiguration unstated?

Full ranking →
1F2LLM v2 8B70.52F2LLM v2 14B70.43F2LLM v2 4B69.44F2LLM v2 1.7B68.65Jina Embeddings v5 Text Small67.5
···
26Vietnamese Bi Encoder57.227Bge Multilingual Gemma256.128Static Similarity Mrl Multilingual v155

Alternatives?

ModelScoreCostPriceWhat changes
CheaperJina Embeddings v5 Text SmallunstatedOpen+20%-55%$0.05iBetterMultilingual E5 Large InstructunstatedOpen+18%-91%$0.01Max input: 8K tokens → 1K tokenscOpenF2LLM v2 8Bunstated+26%——iCheaperE5 Mistral 7B InstructunstatedOpen+18%-80%$0.022CheaperBge M3unstatedOpen+15%-91%$0.01

Run it yourself

estimate
Fits on✓ 8 GB (phone, laptop) · compressed✓ 16 GB · 8-bit✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
Licencehttps://ai.google.dev/gemma/termscommercial use with conditions

Benchmarks

in Embeddings & Retrieval
BEIR · Mean over tasks · English59.2#15
BEIR · Mean over task types · English59.2#15
BEIR · Retrieval · English59.2#15
MTEB Mandarin Chinese v1 · STS · Mandarin Chinese55.2#31
MTEB Mandarin Chinese v1 · Clustering · Mandarin Chinese59.3#35
MTEB Mandarin Chinese v1 · Reranking · Mandarin Chinese68.3#19
MTEB Mandarin Chinese v1 · Retrieval · Mandarin Chinese73.7#24
MTEB Mandarin Chinese v1 · Pair Classification · Mandarin Chinese79.3#28
MTEB German v1 · Reranking · German26.5#78
MTEB French v1 · Reranking · French85.2#11
MTEB French v1 · Pair Classification · French60.8#64
MTEB French v1 · Summarization · French31.3#26
MTEB Indic v1 · Clustering · multilingual (21 languages)34.6#48
MTEB Korean v2 · Mean over tasks · multilingual (2 languages)67.4#15
MTEB Korean v2 · Mean over task types · multilingual (2 languages)62.6#20
MTEB Korean v2 · Retrieval · multilingual (2 languages)70.1#25
MTEB Korean v2 · Clustering · Korean60.1#27
MTEB Korean v2 · STS · Korean80.1#32
MTEB Korean v2 · Classification · Korean62.7#14
MTEB Korean v2 · Pair Classification · Korean65.7#12
MTEB Korean v2 · Reranking · Korean36.8#56
MTEB Polish v1 · Clustering · Polish46.4#17
MTEB Polish v1 · STS · Polish70.6#8
MTEB Polish v1 · Pair Classification · Polish89.8#2
MTEB Medical v1 · Reranking · Mandarin Chinese85.6#34
MTEB Spanish v1 · Reranking · Spanish31.5#93
MTEB Thai v1 · Clustering · Thai39.6#75
VN-MTEB Vietnamese v1 · Reranking · Vietnamese56.1#27
VN-MTEB Vietnamese v1 · Pair Classification · Vietnamese82.5#16
VN-MTEB Vietnamese v1 · Retrieval · Vietnamese20.5#23
VN-MTEB Vietnamese v1 · Clustering · Vietnamese40.1#24
VN-MTEB Vietnamese v1 · STS · Vietnamese66.1#28

FAQ

How much does Bge Multilingual Gemma2 cost?›

$0.11 per 1M tokens.

Is Bge Multilingual Gemma2 free to use?›

There is no free API; it costs $0.11 per 1M tokens. The weights are open, so you can run it yourself.

How good is Bge Multilingual Gemma2, and what is it best at?›

#60 of 238 in Embeddings.

Is Bge Multilingual Gemma2 better than Vietnamese Bi Encoder?›

On VN-MTEB Vietnamese v1 · Reranking · Vietnamese, Bge Multilingual Gemma2 scores 56.1 vs 57.2 — lower.

When was Bge Multilingual Gemma2 released? Is there a newer version?›

Released 2024-07-25. It is the current version in its line.

Is Bge Multilingual Gemma2 open source?›

Yes, the weights are open (https://ai.google.dev/gemma/terms licence).

Can I run Bge Multilingual Gemma2 locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

What is Bge Multilingual Gemma2?›

Bge Multilingual Gemma2 is an open-weights embeddings & retrieval model from BAAI, released Jul 2024.

What are the best alternatives to Bge Multilingual Gemma2?›

Jina Embeddings v5 Text Small (cheaper), Multilingual E5 Large Instruct (better), F2LLM v2 8B (open weights).