Compare rerank models and API pricing
Put the most relevant documents first.
Find rerank models that put the documents a search found in order of relevance to a query, best first, for retrieval-augmented generation, search and recommendations. Every model uses the same rerank endpoint.
5 rerank models are listed today. All of them use one OneInfer API key and pay-as-you-go wallet.
Choose and call a rerank model in 4 steps
The list ranks them by their input price per million tokens. Different billing units are never mixed, so the order is fair.
Lowest comparable price: Qwen: Qwen3 Reranker 0.6B
Currently the lowest-priced option in the provider price per one million input tokens group. Different billing units are never mixed.
$0.011 / 1M input tokens
View pricing and providersAvailable rerank models
Comparable models are ordered by their current normalized price. Pricing content refreshed November 9, 2026.
5 models
| Model |
|---|
Qwen: Qwen3 Reranker 0.6BLowest pricedeepinfra Qwen3 Reranker 0.6B is the smallest Qwen3 reranker: a fast relevance score for each document against a query.
|
deepinfra Llama Nemotron Rerank VL 1B v2 is NVIDIA's compact reranker for documents with charts and tables.
|
deepinfra Qwen3 Reranker 4B scores how relevant each document is to a query, in 100+ languages, and takes an instruction saying what counts as relevant.
|
deepinfra Qwen3 Reranker 8B is the largest of Alibaba's Qwen3 rerankers: it scores how relevant each document is to a query, in 100+ languages, and takes an instruction saying what counts as relevant.
|
deepinfra Nemotron 3.5 Rerank 8B is NVIDIA's reranker for reordering the candidates an embedding search found. Its scores are unbounded; higher is more relevant.
|
How to choose
- RAG
- Search ranking
- Recommendations
- Question answering
Pricing methodology
OneInfer compares only positive prices with the same billing unit. Per-minute, per-character, per-token, per-image, per-video, and per-second rates remain separate. Prices can change, so the current model page and console remain the source of truth.
Frequently asked questions
How are rerank model prices compared?
Rerank models are compared by their positive price per one million input tokens. The query and every document count as input; there is no output to price.
How do I call a rerank model?
Send POST /v1/ula/rerank with the model id, a query and 1 to 1,000 documents. The request follows Cohere’s rerank format. Results come back best first, each with its document’s index and a relevance score; top_n returns only the best few.
Are rerank scores between 0 and 1?
Not always, and scores are not comparable between models. Nemotron 3.5 Rerank returns unbounded scores (logits), where higher means more relevant. Rely on the order of the results rather than a fixed score threshold.