Cohere Embed 5 and Rerank 5 Are Now Live on OpenRouter

According to Cohere Blog:

How It Works

Building a modern RAG pipeline usually requires juggling multiple tools, but OpenRouter now hosts both of Cohere's latest retrieval powerhouses under one roof. The setup uses Embed 5 Pro to build a searchable document index at the highest quality, while Embed 5 Fast handles real-time queries with minimal latency, leveraging a shared embedding space to preserve retrieval accuracy.

Precision Reranking

After fetching initial candidates via vector similarity, Rerank 5 steps in to score and filter the results. Because its relevance scores are fully calibrated across different queries and domains, developers can rely on a single threshold to cleanly strip out weak context before it ever reaches the language model.

Practical Benefits

This unified approach simplifies development and cuts down on infrastructure overhead. Every component, from vector generation to the final chat response, runs through the same API, making it simple to swap models or scale up with enterprise documents.