Cohere AI Updates: November 5, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. Cohere’s Embed 5 and Rerank 5 Arrive on OpenRouter, Where Fast Queries Against a Pro Index Keep 98.4% of All-Pro Quality
Cohere made its Embed 5 and Rerank 5 retrieval models available on OpenRouter as cohere/embed-5-pro, cohere/embed-5-fast, cohere/rerank-5-pro and cohere/rerank-5-fast, served through OpenRouter’s Embeddings and Rerank APIs with the same API key and credits as other OpenRouter models and no extra SDK. On OpenRouter, Embed 5 accepts images and mixed inputs such as PDF pages as image_url content parts, but the embeddings API returns floats only, so the int8 and binary output types stay on Cohere’s own API, even though Cohere calls 1,024-dimension int8 vectors the best quality-to-cost point. Because Pro and Fast share one embedding space at the same dimension, teams can index with Pro and query with Fast: across 40 datasets, Cohere measured that pairing at 98.4% of all-Pro retrieval quality, compared with 96.6% for Fast on both sides. Rerank 5 is text-only, reads up to 32k tokens of query and chunk combined (longer inputs are split into several passes), and returns calibrated scores, so one minimum-score threshold can be applied across queries and domains. The post gives no per-token or per-search prices, which are listed on each model’s OpenRouter page. Source