Cohere Embed 5 and Rerank 5 are on OpenRouter

Cohere put its Embed 5 and Rerank 5 model on OpenRouter, a gateway that gives access to many AI models with one API key (a code that gives access). There are 4 models: Embed 5 Pro, Embed 5 Fast, Rerank 5 Pro and Rerank 5 Fast. A team that makes search or RAG (the model reads found documents first) can add them with no new account. The text gives no test numbers.

How it works

Embed models change text, images or PDF pages into numbers that a search can compare. Rerank models put found results in a sequence of fit to a query. Embed 5 Pro and Fast have the same embedding space, and a team can index with Pro and use Fast for queries.

The numbers

There are no test numbers. The text tells only that Rerank 5 Fast gives Rerank 5 quality for low cost and that it is better than Rerank 4 Pro. Prices are on the page of each model.

What is not known

The text gives no price, no speed and no test on your data. It tells that the OpenRouter models are different from the Cohere API, but it does not tell how.

Insight: Cohere moves from its own API to a shared gateway

In October, Cohere put Rerank 5 on the Cohere API. At this time the same models are on OpenRouter, where one key gives access to many models. In the same month, NVIDIA told of open Nemotron 3.5 search models that cut agent cost by 14% in its tests. This reads as a race to be the first model that a builder tests, because a key that works makes a trial easy. Test 2 search models on your data.