Release Notes

Cohere’s Embed 5 and Rerank 5 models are now available on OpenRouter. If you build on OpenRouter, you can now add Cohere embeddings and reranking to your search and RAG pipelines with your existing OpenRouter API key.

Available models

• Embed 5 Pro (cohere/embed-5-pro): Optimized for the highest retrieval quality, particularly for offline indexing and quality-critical retrieval.
• Embed 5 Fast (cohere/embed-5-fast): Optimized for low latency and high throughput, particularly for interactive search, agent loops, and high-volume query traffic.
• Rerank 5 Pro (cohere/rerank-5-pro): Optimized for the highest ranking quality, particularly for complex, reasoning-heavy queries and long or domain-specific documents.
• Rerank 5 Fast (cohere/rerank-5-fast): Rerank 5 quality for latency- and cost-sensitive workloads. Outperforms Rerank 4 Pro.

Embed 5 accepts text, images, and mixed text-and-image inputs such as PDF pages. Embed 5 Pro and Fast share an embedding space, so you can index your corpus with Pro and embed queries with Fast.

For pricing, see each model’s OpenRouter page.

Getting started

See Using Cohere Models on OpenRouter for the full list of Cohere models on OpenRouter, a setup checklist, and the differences from the Cohere API. For an end-to-end retrieval pipeline, see OpenRouter’s RAG with Embeddings & Rerank cookbook.

For more on the models themselves, see the Embed 5 announcement and the Rerank 5 announcement.