Llama Nemotron Rerank VL 1B v2
Llama Nemotron Rerank VL 1B v2 is a vision-language reranking model from Nvidia with a 10,240-token context window.
Nvidia
DeepInfraMultimodal reranking model for retrieval pipelines
Llama Nemotron Rerank VL 1B v2 is a reranking model developed by Nvidia, built on the Llama architecture and released in March 2026. It is designed to score and reorder candidate documents or passages retrieved in a retrieval-augmented generation (RAG) pipeline, improving the relevance of results passed to a downstream language model. The model supports vision-language inputs, meaning it can process both text and image content when determining relevance. It is served through DeepInfra and carries a 1-billion parameter footprint, making it relatively compact for a multimodal reranker.
This model fits into the Nemotron family of models that Nvidia has developed for enterprise retrieval and search workflows. Its primary use case is as a second-stage ranker that takes an initial set of retrieved candidates and reorders them by relevance score before generation occurs. The 10,240-token context window allows it to handle moderately long documents within a single pass. Developers building multimodal RAG systems who need a lightweight reranker capable of handling image-bearing documents are the intended audience.
What Llama Nemotron Rerank VL 1B v2 supports
Document Reranking
Scores and reorders retrieved candidate passages by relevance, functioning as a second-stage ranker in RAG pipelines to improve result quality before generation.
Vision-Language Input
Processes both text and image content together when scoring document relevance, enabling reranking of multimodal documents that contain visual elements.
Long Context Handling
Supports a context window of 10,240 tokens, allowing moderately long documents or multi-turn query-document pairs to be evaluated in a single pass.
RAG Pipeline Integration
Designed to slot into retrieval-augmented generation workflows as a reranking stage, compatible with standard retrieval APIs and vector search outputs.
Compact Model Footprint
At 1 billion parameters, the model is sized for efficient inference, reducing latency and compute cost in production reranking deployments.
Ready to build with Llama Nemotron Rerank VL 1B v2?
Get Started FreeCommon questions about Llama Nemotron Rerank VL 1B v2
What is the context window for Llama Nemotron Rerank VL 1B v2?
The model supports a context window of 10,240 tokens, which determines the maximum combined length of a query and candidate document it can evaluate in a single reranking call.
What does a reranking model do, and how does this one fit into a pipeline?
A reranking model takes a set of candidate documents already retrieved by a first-stage retriever and assigns relevance scores to reorder them. Llama Nemotron Rerank VL 1B v2 acts as that second-stage component, returning ranked results that a downstream language model can then use for generation.
Does this model support image inputs?
Yes. The 'VL' in the model name stands for vision-language, indicating it can process both text and image content when evaluating document relevance, making it suitable for multimodal retrieval scenarios.
What is the pricing for this model?
Pricing information is not published in the available metadata. You should check the DeepInfra provider page for current per-token or per-request pricing details.
Who publishes this model and when was it released?
Llama Nemotron Rerank VL 1B v2 is published by Nvidia and was released in March 2026. It is served via DeepInfra under the model identifier nvidia/llama-nemotron-rerank-vl-1b-v2.
Explore similar models
NvidiaNemotron 3.5 Lightning
Nemotron 3.5 Lightning is a text generation model from Nvidia designed for fast reasoning with a 262,144 token context window.
NvidiaNemotron 3 Ultra 550B
Nemotron 3 Ultra 550B is a text generation model from Nvidia with a 262,144-token context window and reasoning support.
NvidiaNemotron 3 Nano 30B
Nemotron 3 Nano 30B is a text generation model from Nvidia with a 1,000,000-token context window and mixture-of-experts architecture.
CohereCohere Rerank 4 Fast
Cohere Rerank 4 Fast is a reranking model from Cohere designed to score and reorder search results with a 4096 token context window.
Voyage AIVoyage Rerank 2.5
Voyage Rerank 2.5 is a reranking model from Voyage AI designed to improve document retrieval relevance with a 32,000-token context window.
QwenQwen3 Reranker 0.6B
Qwen3 Reranker 0.6B is a lightweight reranking model from Qwen designed to reorder retrieved documents with a 32,768-token context window.
Start building with Llama Nemotron Rerank VL 1B v2
No API keys required. Create AI-powered workflows with Llama Nemotron Rerank VL 1B v2 in minutes — free.