Cohere Releases Embed 5: How It Compares to Voyage 4 Large, Gemini Embedding 2, and OpenAI
Cohere launched Embed 5 Pro and Fast, multimodal embeddings with a 128K context for search, RAG, and agents.
Cohere released Embed 5 Pro and Embed 5 Fast, multimodal embedding models for enterprise search, RAG, and agentic retrieval. Both accept text, images, and fused inputs, cover more than 100 languages, read up to 128K tokens, and share one embedding space with outputs from 256 to 2048 dimensions in float, int8, or binary. Cohere reports Pro at 85.8 and Fast at 84.5 on ViDoRe V3 using its RCP-nDCG@10 metric, ahead of Voyage 4 Large (83.7), Gemini Embedding 2 (83.2), and OpenAI text-embedding-3-large (75.5); independent replication is still pending. Both tiers are generally available on the Cohere API, Model Vault, Microsoft Foundry, and Amazon SageMaker, with private serving through vLLM.