Nemotron Colembed VL 4B v2
Jan 2026
Nemotron Colembed VL 4B v2Compare inon
Nemotron Colembed VL 4B v2 measured on 5 of 203 Multimodal benchmarks5
BenchmarkByScoreRank
No benchmark matches
self-host onlyprice
88.6MIEB lite
Jan 2026released
2,560Dimensions
262K tokensMax input
Ranking?
Full ranking →···
12Tomoro Colqwen3 Embed 8B61.613Nemotron Colembed VL 4B v261.514pplx Embed v2 Late 0.6B61.5Strong at
Alternatives?
ModelScoreCostPriceWhat changes
BetterEVIE 8BOpen+8%——Memory (4-bit): 2.9 GB → 5 GBBetterEVIE 4.5BOpen+7%——wBetterWebAI ColVec1.1 8BOpen+6%——Memory (4-bit): 2.9 GB → 5 GB
Betterpplx Embed v2 Late 9BOpen+5%——Max input: 262K tokens → 4K tokens+1vBetterVultronRetrieverPrime Qwen3.5 8BOpen+4%——Memory (4-bit): 2.9 GB → 5 GBModelScoreCostPriceWhat changes
Nothing here yet.
ModelScoreCostPriceWhat changes
EVIE 8BOpen+8%——Memory (4-bit): 2.9 GB → 5 GBEVIE 4.5BOpen+7%——wWebAI ColVec1.1 8BOpen+6%——Memory (4-bit): 2.9 GB → 5 GBpplx Embed v2 Late 9BOpen+5%——Max input: 262K tokens → 4K tokens+1vVultronRetrieverPrime Qwen3.5 8BOpen+4%——Memory (4-bit): 2.9 GB → 5 GBModelScoreCostPriceWhat changes
EVIE 8BOpen+8%——Memory (4-bit): 2.9 GB → 5 GBEVIE 4.5BOpen+7%——wWebAI ColVec1.1 8BOpen+6%——Memory (4-bit): 2.9 GB → 5 GBpplx Embed v2 Late 9BOpen+5%——Max input: 262K tokens → 4K tokens+1vVultronRetrieverPrime Qwen3.5 8BOpen+4%——Memory (4-bit): 2.9 GB → 5 GBModelScoreCostPriceWhat changes
Nemotron Colembed VL 8B v2Open+3%——Memory (4-bit): 2.9 GB → 5.2 GBNemotron 3.5 Embed 8B BF16Open-2%—$0.035Max input: 262K tokens → 33K tokens+1Llama Nemotron Colembed VL 3B v2Open-3%——Max input: 262K tokens → 8K tokensNemotron 3.5 Embed 1B BF16Open-6%—$0.015Max input: 262K tokens → 33K tokensLlama Nemoretriever Colembed 3B v1Open-7%——Max input: 262K tokens → 8K tokensLlama Nemoretriever Colembed 1B v1Open-10%——Max input: 262K tokens → 8K tokensLlama Nemotron Embed VL 1B v2Openn/a—$0.01Max input: 262K tokens → 10K tokensOmni Embed Nemotron 3BOpenn/a——Max input: 262K tokens → 33K tokensAll NVIDIA models →Run it yourself
estimateFits 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/nemotron-colembed-vl-4b-v2/blob/main/LICENSEno commercial use
Benchmarks
in Embeddings & RetrievalFAQ
How much does Nemotron Colembed VL 4B v2 cost?›
There is no per-unit API price: self-host only.
Is Nemotron Colembed VL 4B v2 free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is Nemotron Colembed VL 4B v2, and what is it best at?›
#7 of 89 in Multimodal. Strongest at: Embeddings & Retrieval.
Is Nemotron Colembed VL 4B v2 better than Tomoro Colqwen3 Embed 8B?›
On ViDoRe v3 · Mean over tasks · multilingual (6 languages), Nemotron Colembed VL 4B v2 scores 61.5 vs 61.6 — about the same.
When was Nemotron Colembed VL 4B v2 released? Is there a newer version?›
Released 2026-01-07. It is the current version in its line.
Is Nemotron Colembed VL 4B v2 open source?›
Yes, the weights are open (https://huggingface.co/nvidia/nemotron-colembed-vl-4b-v2/blob/main/LICENSE licence).
Can I run Nemotron Colembed VL 4B v2 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
What is Nemotron Colembed VL 4B v2?›
Nemotron Colembed VL 4B v2 is an open-weights embeddings & retrieval model from NVIDIA, released Jan 2026.
What are the best alternatives to Nemotron Colembed VL 4B v2?›
EVIE 8B (better).