What it does
Cohere builds language and retrieval models for organisations that want to keep their data and infrastructure under their own control. Its line-up is Command for generation, with agentic and tool-orchestration abilities, multimodal input and support for 49 languages; Transcribe for speech-to-text across 14 languages; Embed for turning text into vectors for search and comparison; and Rerank for ordering results by relevance. Models can be deployed inside your own virtual private cloud, on-premises, or on a dedicated Cohere-managed Model Vault, and can be customised with proprietary data. The site also presents North, a workplace product that connects an organisation's tools, and its news section announces North Automations for workflow orchestration. Developers can get an API key, read the documentation and try the models in a Playground.
Capabilities
- Command generative models with agentic, tool-use and multimodal abilities in 49 languages
- Transcribe speech-to-text covering 14 languages
- Embed and Rerank models for semantic search and relevance ranking
- Deployment in a VPC, on-premises or on a managed Model Vault
- North workplace product that connects an organisation's tools
When to use it
- You need language models that can run inside your own cloud or on-premises
- You are building search or retrieval and need embedding and reranking models
- You want speech transcription that plugs into generative and retrieval systems
