What changed
Cohere released Embed 5 Pro and Fast for text, image and fused retrieval across more than 100 languages and contexts up to 128K tokens. Both models share an embedding space, allowing teams to index with Pro and query with Fast without rebuilding the index, and support Matryoshka dimensions plus float, int8 and binary outputs.
Why it matters
A compatible quality and latency tier can reduce query-path cost without forcing a second corpus index. The multimodal and compressed-output options also make storage, retrieval quality and private-serving requirements explicit engineering choices rather than fixed model properties.
What to do
Build a frozen evaluation set from your own corpus, index one copy with Pro, query it with both tiers at equal dimensions, and compare recall, latency and total storage across float, int8 and binary outputs before replacing an existing index.
Caveat
Performance figures are Cohere-run, and some evaluation sets or annotations are internal or vendor-curated. Shared-space compatibility still requires matching output dimensions, private deployment has separate infrastructure requirements, and benchmark gains do not establish application answer quality.