Reranking
Models that reorder search results by relevance to a query.
The cross-encoder models that rerank retrieved passages against a query, the single biggest lever on retrieval quality. Each entry shows how many documents it can score at once and how much text it reads, with a per-million-token price. The vector database applies one by default. Name a specific model when accuracy or language coverage matters.
9 models
Cohere Rerank 3.5RerankingCohere
$0.0020/ search
1,000
Max docs
4.1K tok
Context
Cohere Rerank 4 FastRerankingCohere
$0.0010/ search
1,000
Max docs
4.1K tok
Context
Cohere Rerank 4 ProRerankingCohere
$0.0020/ search
1,000
Max docs
4.1K tok
Context
Llama Nemotron Rerank VL 1B v2RerankingNvidia· via
DeepInfra
$0.010/ 1M tokens
1,000
Max docs
10.2K tok
Context
Qwen3 Reranker 0.6BRerankingQwen· via
DeepInfra
$0.010/ 1M tokens
1,000
Max docs
32.8K tok
Context
Qwen3 Reranker 4BRerankingQwen· via
DeepInfra
$0.025/ 1M tokens
1,000
Max docs
32.8K tok
Context
Qwen3 Reranker 8BRerankingQwen· via
DeepInfra
$0.050/ 1M tokens
1,000
Max docs
32.8K tok
Context
Voyage Rerank 2.5RerankingVoyage AI
$0.050/ 1M tokens
1,000
Max docs
32K tok
Context
Voyage Rerank 2.5 LiteRerankingVoyage AI
$0.020/ 1M tokens
1,000
Max docs
32K tok
Context