RELEASE · MODELS · #1269
Cohere releases Embed 5 embeddings family with Pro and Fast tiers
Cohere announced Embed 5, a new family of embeddings offered in Pro and Fast tiers that support multimodal inputs, 100+ languages, and a 128K-token context window. Cohere says Embed 5 Pro achieves the highest average scores in their tests (notably on ViDoRe V3 and financial benchmarks), while Embed 5 Fast offers lower-latency, lower-cost retrieval; both share a single embedding space and are available via the Cohere API, Model Vault, Microsoft Foundry, and Amazon SageMaker with pricing at $0.12/million tokens (Pro) and $0.08/million tokens (Fast).
KEY POINTS
- Cohere announced Embed 5, a new family of embeddings offered in Pro and Fast tiers that support multimodal inputs, 100+ languages, and a 128K-token context window.
- Cohere says Embed 5 Pro achieves the highest average scores in their tests (notably on ViDoRe V3 and financial benchmarks), while Embed 5 Fast offers lower-latency, lower-cost retrieval; both share a single embedding space and are available via the Cohere API, Model Vault, Microsoft Foundry, and Amazon SageMaker with pricing at $0.12/million tokens (Pro) and $0.08/million tokens (Fast).
- New high-quality, multimodal enterprise embeddings with shared Pro/Fast index compatibility and large context support can materially improve retrieval, RAG, and search pipelines while controlling inference costs.
WHY IT MATTERS
New high-quality, multimodal enterprise embeddings with shared Pro/Fast index compatibility and large context support can materially improve retrieval, RAG, and search pipelines while controlling inference costs.