Cohere

About Cohere

Cohere sells AI to enterprises rather than to consumers, and it competes on where the model runs as much as on the model itself. Its pricing page (read 16 October 2026) lists four products: North, described as an all-in-one AI platform for automating tasks safely across an enterprise; Command, its generative language models; Embed, a semantic representation model for search; and Rerank, for retrieval optimisation. API calls made from a trial key are free, but trial keys are rate-limited and not permitted for production or commercial use. The per-token prices printed on the page are labelled as legacy models for existing customers — Command at $1.00 per million input tokens and $2.00 per million output, Command R+ 08-2024 at $2.50 and $10.00, Command-light at $0.30 and $0.60, Aya Expanse at $0.50 and $1.50 — and Rerank is billed per search unit, one query with up to 100 documents, without a figure shown. Current models and enterprise work are priced on request, including private deployment on-premises or in an isolated VPC and bespoke model customisation. That positions it clearly: for regulated organisations that want retrieval-augmented search and generation without sending data to a shared cloud. The limits: no published price for the current line-up, legacy rates that may not apply to a new contract, and a developer-first product that needs engineering time before anyone else benefits. On this site, Claude Code and GitHub Copilot are coding assistants rather than model platforms, and Glean is the packaged enterprise-search option for teams that would rather buy than build. Verdict: a strong choice for enterprises building search or assistants on private infrastructure — start with a trial key, and get a quote before you set a budget.