1B- input
- $0.00
- output
- $0.00
- Context
- 10Ktokens
- Released
- 10mo ago
- Size
- 1Bparameters
Cost/1M tokens
Developed by NVIDIA, llama-nemotron-rerank-vl-1b-v2 is a fine-tune of the company's Eagle family, combining a SigLIP 2 400M vision encoder with a Llama 3.2 1B language model. NVIDIA says its distinguishing feature is reranking document pages supplied as images, text, or both — including pages containing text, tables, charts, and infographics — for relevance to a text query. It is built for reranking candidate documents in multimodal retrieval systems such as vision RAG pipelines.
How it compares
Additional source evidence
- Source model
- nvidia/llama-nemotron-rerank-vl-1b-v2
- Source family
- nemotron
- Source type
- reranking
- Last updated
- 2025-12-18
- Model context
- 10K
- Model output limit
- 4K
- Open weights
- Yes
- File attachments
- Yes
- Reasoning
- No
- Tool calling
- No
- Temperature control
- No
Complete source record (JSON)
{
"id": "nvidia/llama-nemotron-rerank-vl-1b-v2",
"name": "Llama Nemotron Rerank VL 1B v2",
"description": "Reranking model for improving retrieval quality in search and recommendation systems",
"family": "nemotron",
"type": "reranking",
"last_updated": "2025-12-18",
"attachment": true,
"reasoning": false,
"tool_call": false,
"temperature": false,
"modalities": {
"input": [
"text",
"image"
],
"output": [
"text"
]
},
"limit": {
"context": 10240,
"output": 4096
},
"release_date": "2025-12-18",
"open_weights": true
}Frequently asked questions
Running locally
Estimated 1.2 GB at Q4, 1.7 GB at Q8 and 2.8 GB at FP16. These figures are estimates derived from the 1B parameter count, with weights plus about 15% and 0.5 GB for a short context.
Yes. It fits at FP16 with an estimated 2.8 GB needed, so a 24 GB, 16 GB, 12 GB or 8 GB GPU is enough.
Yes. At Q4 it needs an estimated 1.2 GB, and assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits is 8 GB.
Privacy and data
Yes. The model has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 10K tokens (10,240) and the maximum output is 9K tokens (9,216).
It supports text and image inputs. Audio, video and PDF inputs are not supported.
No. Tool calling (function calling) is not supported.
No. A reasoning or thinking mode is not supported.
Quality and benchmarks
Closest alternatives are Rerank 4 Fast by Cohere and Rerank 4 Pro by Cohere. They were matched on model kind and release date, with no comparable price.
Rerank 4 FastLlama Nemotron Rerank VL 1B V2 vs Rerank 4 FastRerank 4 ProLlama Nemotron Rerank VL 1B V2 vs Rerank 4 Pro
Licensing and access
Llama Nemotron Rerank VL 1B V2 has open weights.
The weights are available on Hugging Face in the repository nvidia/llama-nemotron-rerank-vl-1b-v2.
Basics
NVIDIA made Llama Nemotron Rerank VL 1B V2, and it was released on December 18, 2025.
Yes. Llama Nemotron Rerank VL 1B V2 has open weights.
Generated from Vertical Tabs data, updated November 3, 2026.
Discussion
No comments yet
Be the first to share your thoughts.