H

KaLM Embedding Multilingual Mini Instruct v1

Oct 2024
HKaLM Embedding Multilingual Mini Instruct v1
Compare in
on
self-host onlyprice
44.4MTEB English v2
Oct 2024released
896Dimensions
33K tokensMax input

Rankingconfiguration unstated?

Full ranking →
1Octen Law 8B v176.82Litil Embed 0.6B73.33Dinghy Law 8B v172.64Dinghy Law 4B v171.25Euler Legal Embedding v170.4
···
37Multilingual E5 Large Instruct54.938KaLM Embedding Multilingual Mini Instruct v154.539Gte Multilingual Base54.5

Strong at

C
Embeddings & Retrieval#131 of 200

Alternatives?

ModelScoreCostPriceWhat changes
VBetterVoyage 3 Liteunstated+11%—$0.02Max input: 33K tokens → 32K tokenslOpenOcten Law 8B v1+41%——Memory (4-bit): 0.3 GB → 4.5 GBBetterJina Embeddings v3unstatedOpen+4%—$0.05Max input: 33K tokens → 8K tokensVBetterVoyage 3unstated+18%—$0.06Max input: 33K tokens → 32K tokensVBetterVoyage Law 2unstated+20%—$0.12Max input: 33K tokens → 16K tokens

Run it yourself

estimate
Fits on✓ 8 GB (phone, laptop) · full quality✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceMITfree for commercial use

Benchmarks

in Embeddings & Retrieval
BEIR · Mean over tasks · English51.9#79
BEIR · Mean over task types · English51.9#79
BEIR · Retrieval · English51.9#79
BEIR-NL · Mean over tasks · Dutch26.1#11
BEIR-NL · Mean over task types · Dutch26.1#11
BEIR-NL · Retrieval · Dutch26.1#11
MTEB Mandarin Chinese v1 · Retrieval · Mandarin Chinese70.5#44
MTEB Mandarin Chinese v1 · STS · Mandarin Chinese47.9#50
MTEB Mandarin Chinese v1 · Reranking · Mandarin Chinese64.4#46
MTEB Mandarin Chinese v1 · Clustering · Mandarin Chinese50.9#43
MTEB Mandarin Chinese v1 · Pair Classification · Mandarin Chinese73.1#39
MTEB Persian v2 · Reranking · Persian55.3#56
JMTEB v2 · Clustering · Japanese45.8#52
JMTEB v2 · STS · Japanese79.2#50
LongEmbed · Mean over tasks · English65.1#29
LongEmbed · Mean over task types · English65.1#29
LongEmbed · Retrieval · English65.1#29
MTEB German v1 · Clustering · German34.6#25
MTEB German v1 · Retrieval · German43.7#34
MTEB German v1 · Classification · German66.8#35
MTEB German v1 · Pair Classification · German53.3#40
MTEB German v1 · Reranking · German37.6#53
MTEB Europe v1 · Retrieval · multilingual (25 languages)50.5#40
MTEB Europe v1 · Bitext Mining · multilingual (28 languages)82.4#70
MTEB Europe v1 · Classification · multilingual (23 languages)59#26
MTEB Europe v1 · Clustering · multilingual (29 languages)40.9#58
MTEB Europe v1 · Instruction Reranking · English-2.6#150
MTEB Europe v1 · Pair Classification · multilingual (11 languages)87.4#36
MTEB Europe v1 · Reranking · multilingual (12 languages)57.3#58
MTEB Europe v1 · Multilabel Classification · multilingual (23 languages)18.4#38
MTEB French v1 · Retrieval · multilingual (2 languages)53.8#26
MTEB French v1 · Reranking · French80.8#21
MTEB French v1 · Classification · French67.7#28
MTEB French v1 · Summarization · French30.9#40
MTEB French v1 · Pair Classification · French61#59
MTEB Indic v1 · Mean over tasks · multilingual (25 languages)57.2#41
MTEB Indic v1 · Mean over task types · multilingual (25 languages)51.5#42
MTEB Indic v1 · Classification · multilingual (8 languages)61.3#34
MTEB Indic v1 · Retrieval · multilingual (14 languages)60#53
MTEB Indic v1 · Bitext Mining · multilingual (20 languages)49.6#42
MTEB Indic v1 · Reranking · multilingual (2 languages)76.8#84
MTEB Indic v1 · Pair Classification · Hindi67.8#69
MTEB Indic v1 · STS · multilingual (13 languages)23.1#76
MTEB Indic v1 · Clustering · multilingual (21 languages)22.3#87
MTEB Korean v2 · Reranking · Korean27.7#63
MTEB Korean v2 · Classification · Korean55.4#41
MTEB Dutch v1 · Mean over tasks · Dutch53.6#42
MTEB Dutch v1 · Mean over task types · Dutch60.5#35
MTEB Dutch v1 · Classification · Dutch60.5#37
MTEB Dutch v1 · Clustering · Dutch37#30
MTEB Dutch v1 · Retrieval · Dutch52.1#57
MTEB Dutch v1 · STS · Dutch74.7#53
MTEB Dutch v1 · Multilabel Classification · Dutch36.3#29
MTEB Dutch v1 · Pair Classification · Dutch76.4#42
MTEB Dutch v1 · Reranking · Dutch86.5#75
MTEB Polish v1 · Pair Classification · Polish80.6#47
MTEB Polish v1 · Classification · Polish63.2#23
MTEB Polish v1 · STS · Polish67.3#25
MTEB Portuguese v1 · Multilabel Classification · Portuguese20.4#69
MTEB Law v1 · Mean over tasks · multilingual (3 languages)54.5#38
MTEB Law v1 · Mean over task types · multilingual (3 languages)54.5#38
MTEB Law v1 · Retrieval · multilingual (3 languages)54.5#38
MTEB Medical v1 · Mean over tasks · multilingual (10 languages)57.3#25
MTEB Medical v1 · Mean over task types · multilingual (10 languages)58.8#23
MTEB Medical v1 · Retrieval · multilingual (10 languages)59.7#28
MTEB Medical v1 · Clustering · English34.2#141
MTEB Medical v1 · Reranking · Mandarin Chinese82.6#43
FollowIR · Mean over tasks · English-2.6#150
FollowIR · Mean over task types · English-2.6#150
FollowIR · Instruction Reranking · English-2.6#150
MTEB Spanish v1 · Pair Classification · Spanish68.9#43
MTEB Spanish v1 · Reranking · Spanish48.8#54
MTEB Thai v1 · Clustering · Thai38.3#83
MTEB Thai v1 · Pair Classification · Thai67.6#72
MTEB Scandinavian v1 · Clustering · multilingual (2 languages)39.9#50
MTEB Scandinavian v1 · Classification · multilingual (6 languages)56.1#58
MTEB Scandinavian v1 · Bitext Mining · multilingual (3 languages)66.2#82
MTEB Russian v1.1 · Mean over tasks · Russian60#27
MTEB Russian v1.1 · Mean over task types · Russian57.3#30
MTEB Russian v1.1 · Classification · Russian63.4#32
MTEB Russian v1.1 · STS · Russian71.2#42
MTEB Russian v1.1 · Reranking · Russian54.7#59
MTEB Russian v1.1 · Clustering · Russian55.3#40
MTEB Russian v1.1 · Pair Classification · Russian55.6#85
MTEB Russian v1.1 · Retrieval · Russian64.4#29
MTEB Russian v1.1 · Multilabel Classification · Russian36.5#42
RTEB French beta · Mean over tasks · multilingual (2 languages)29.1#93
RTEB German beta · Mean over tasks · German48.8#61
RTEB Law beta · Mean over tasks · multilingual (4 languages)46#61
MTEB multilingual v2 · Pair Classification · multilingual (24 languages)77.5#62
MTEB multilingual v2 · Retrieval · multilingual (120 languages)50.8#78
MTEB multilingual v2 · Bitext Mining · multilingual (1026 languages)64.2#59
MTEB multilingual v2 · Instruction Reranking · English-2.6#150
MTEB multilingual v2 · Multilabel Classification · multilingual (25 languages)21.6#51
MTEB multilingual v2 · Reranking · multilingual (22 languages)58.2#75

FAQ

How much does KaLM Embedding Multilingual Mini Instruct v1 cost?›

There is no per-unit API price: self-host only.

Is KaLM Embedding Multilingual Mini Instruct v1 free to use?›

There is no free API. The weights are open, so you can run it yourself.

How good is KaLM Embedding Multilingual Mini Instruct v1, and what is it best at?›

#131 of 200 in Embeddings.

Is KaLM Embedding Multilingual Mini Instruct v1 better than Multilingual E5 Large Instruct?›

On MTEB Law v1 · Retrieval · multilingual (3 languages), KaLM Embedding Multilingual Mini Instruct v1 scores 54.5 vs 54.9 — about the same.

When was KaLM Embedding Multilingual Mini Instruct v1 released? Is there a newer version?›

Released 2024-10-23. It is the current version in its line.

Is KaLM Embedding Multilingual Mini Instruct v1 open source?›

Yes, the weights are open (MIT licence).

Can I run KaLM Embedding Multilingual Mini Instruct v1 locally?›

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

What is KaLM Embedding Multilingual Mini Instruct v1?›

KaLM Embedding Multilingual Mini Instruct v1 is an open-weights embeddings & retrieval model from HIT-TMG, released Oct 2024.

What are the best alternatives to KaLM Embedding Multilingual Mini Instruct v1?›

Voyage 3 Lite (better), Octen Law 8B v1 (open weights).