Llama Embed Nemotron 8B

Oct 2025
Llama Embed Nemotron 8B
Compare in
on
self-host onlyprice
79.6MTEB Scandinavian v1
Oct 2025released
4,096Dimensions
131K tokensMax input

Ranking?

Full ranking →
1KURE v281.12KURE v380.33KURE Sparse Base79.74Comsat Embed Ko 8B Preview79.35Qwen 3 Embedding 8B78.4
···
33Ko Embed v059.634Llama Embed Nemotron 8B56.335F2LLM v2 80M55.4

Alternatives?

ModelScoreCostPriceWhat changes
BetterQwen 3 Embedding 8BunstatedOpen+39%—$0.01Max input: 131K tokens → 33K tokensnOpenKURE v2+44%——Max input: 131K tokens → 8K tokensBetterQwen 3 Embedding 4BunstatedOpen+37%—$0.02Max input: 131K tokens → 33K tokensBetterBge Multilingual Gemma2unstatedOpen+25%—$0.11Max input: 131K tokens → 8K tokens+1nBetterKURE v3Open+43%——Max input: 131K tokens → 8K tokens

Run it yourself

estimate
Fits on✓ 8 GB (phone, laptop) · 8-bit✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
Licencehttps://huggingface.co/nvidia/llama-embed-nemotron-8b/blob/main/LICENSEcommercial use with conditions

Benchmarks

in Embeddings & Retrieval
MTEB Europe v1 · Instruction Reranking · English10.8#3
MTEB Europe v1 · Multilabel Classification · multilingual (23 languages)22#13
MTEB Europe v1 · Clustering · multilingual (29 languages)49.9#13
MTEB Europe v1 · Reranking · multilingual (12 languages)62.5#12
MTEB Europe v1 · Bitext Mining · multilingual (28 languages)92.1#7
MTEB French v1 · Pair Classification · French66.8#17
MTEB Indic v1 · STS · multilingual (13 languages)58.2#11
MTEB Indic v1 · Reranking · multilingual (2 languages)88.5#6
MTEB Indic v1 · Pair Classification · Hindi80.1#16
MTEB Indic v1 · Clustering · multilingual (21 languages)57#8
MTEB Korean v2 · Retrieval · multilingual (2 languages)56.3#34
MTEB Dutch v1 · Mean over tasks · Dutch66.2#2
MTEB Dutch v1 · Mean over task types · Dutch70.7#1
MTEB Dutch v1 · Retrieval · Dutch67#4
MTEB Dutch v1 · Classification · Dutch73.4#3
MTEB Dutch v1 · Multilabel Classification · Dutch41.5#4
MTEB Dutch v1 · STS · Dutch86.4#5
MTEB Dutch v1 · Clustering · Dutch50.8#1
MTEB Dutch v1 · Reranking · Dutch91.7#14
MTEB Dutch v1 · Pair Classification · Dutch83.9#1
MTEB Portuguese v1 · Multilabel Classification · Portuguese29#7
FollowIR · Mean over tasks · English10.8#3
FollowIR · Mean over task types · English10.8#3
FollowIR · Instruction Reranking · English10.8#3
MTEB Spanish v1 · Pair Classification · Spanish74.8#13
MTEB Thai v1 · Pair Classification · Thai81.4#15
MTEB Thai v1 · Clustering · Thai63.4#10
MTEB Scandinavian v1 · Bitext Mining · multilingual (3 languages)79.6#14
MTEB Russian v1.1 · Pair Classification · Russian67.2#10
MTEB multilingual v2 · Mean over tasks · multilingual (1037 languages)69.5#5
MTEB multilingual v2 · Mean over task types · multilingual (1037 languages)61.1#5
MTEB multilingual v2 · Clustering · multilingual (208 languages)54.4#15
MTEB multilingual v2 · STS · multilingual (35 languages)79.4#5
MTEB multilingual v2 · Classification · multilingual (119 languages)73.2#6
MTEB multilingual v2 · Retrieval · multilingual (120 languages)68.7#10
MTEB multilingual v2 · Multilabel Classification · multilingual (25 languages)29.9#10
MTEB multilingual v2 · Bitext Mining · multilingual (1026 languages)81.7#5
MTEB multilingual v2 · Instruction Reranking · English10.8#3
MTEB multilingual v2 · Reranking · multilingual (22 languages)67.8#7
MTEB multilingual v2 · Pair Classification · multilingual (24 languages)84#9

FAQ

How much does Llama Embed Nemotron 8B cost?›

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

Is Llama Embed Nemotron 8B free to use?›

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

How good is Llama Embed Nemotron 8B, and what is it best at?›

#11 of 206 in Embeddings.

Is Llama Embed Nemotron 8B better than Ko Embed v0?›

On MTEB Korean v2 · Retrieval · multilingual (2 languages), Llama Embed Nemotron 8B scores 56.3 vs 59.6 — lower.

When was Llama Embed Nemotron 8B released? Is there a newer version?›

Released 2025-10-23. It is the current version in its line.

Is Llama Embed Nemotron 8B open source?›

Yes, the weights are open (https://huggingface.co/nvidia/llama-embed-nemotron-8b/blob/main/LICENSE licence).

Can I run Llama Embed Nemotron 8B locally?›

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

What is Llama Embed Nemotron 8B?›

Llama Embed Nemotron 8B is an open-weights embeddings & retrieval model from NVIDIA, released Oct 2025.

What are the best alternatives to Llama Embed Nemotron 8B?›

Qwen 3 Embedding 8B (better), KURE v2 (open weights).