Cohere is a Toronto-based AI company founded in 2019 by Aidan Gomez (a co-author of the “Attention Is All You Need” paper), Ivan Zhang, and Nick Frosst. Unlike OpenAI or Anthropic, Cohere targets enterprises exclusively — particularly in finance, healthcare, public sector, and regulated industries — building its entire product strategy around data privacy, deployment flexibility, and retrieval-first AI. Its model portfolio is purpose-designed for enterprise tasks: Command models for instruction-following, agentic, and multilingual work; Embed models for semantic search; and Rerank models for retrieval accuracy that consistently outperform competitors in enterprise search benchmarks. North, its AI workspace platform, packages these capabilities for knowledge workers. From $13M ARR in late 2023, the company grew to approximately $70M ARR by early 2025, with major customers including Oracle, Royal Bank of Canada, McKinsey, and STC deploying models in production.
Cohere
Enterprise-grade LLMs, retrieval, and agents with full deployment flexibility
Compliance
HIPAA ISO 27001 SOC 2
Key Features
- Command A (flagship model): The most performant Command model to date, delivering 150% of the throughput of its predecessor on only two GPUs, with a 256K context window, strong agentic and multilingual capabilities, and performance that Cohere states matches or exceeds GPT-4o on agentic enterprise tasks at materially lower compute cost.
- Embed v3.0: A cutting-edge AI search model enhanced with multimodal capabilities, allowing it to generate embeddings from both text and images — used to power hybrid semantic search pipelines across enterprise knowledge bases.
- Rerank 4: The most advanced reranker available, delivering best-in-class retrieval and outperforming MongoDB Voyage and Elasticsearch Jina rerankers in overall search relevance, with self-learning capability that continuously improves precision on domain-specific queries without additional annotated data.
- North (AI workspace): An enterprise AI workspace combining Command, Compass, Embed, and Rerank for knowledge workers to interact with company data through secure AI agents, advanced search, and generative AI — with audit logging, access controls, and compliance features.
- Deployment flexibility: Supports cloud API, private cloud (VPC), on-premises, and air-gapped deployments across AWS, Azure, GCP, and Oracle — a key differentiator for regulated industries where data cannot leave the customer’s infrastructure.
- Model customization and fine-tuning: Enables domain-specific model adaptation for enterprises needing performance on proprietary vocabularies, regulatory language, or internal processes.
Use Cases
- For regulated enterprises requiring on-premises deployment: A financial services firm or government agency deploys Command A on its own infrastructure, processing sensitive documents without data leaving a controlled environment.
- For enterprise search and RAG applications: A company with large internal knowledge bases uses Embed and Rerank to build retrieval pipelines that surface accurate, relevant answers from structured and unstructured documents, with Rerank 4 improving precision over standard vector search alone.
- For multinational organizations needing multilingual AI: A global enterprise deploys Command A for internal knowledge work and customer-facing applications across 23+ supported languages without maintaining separate region-specific models.
Pricing MODELS
Enterprise, Usage-based
Pricing Summary
Usage-based API pricing for cloud access; enterprise is custom-quoted with options for committed usage and volume discounts. On-premises and private cloud deployments are enterprise-only with custom licensing. A free trial is available via the API for evaluation.
Company Size Fit
Enterprise Mid-market SMB
Technical Snapshot
API Available
Yes
LLM Provider
Proprietary
Open Source
Partial
Deployment Options
Cloud Saas, Hybrid, On-premise
Notable Customers
Oracle, Fujitsu, Notion. Public materials also reference Royal Bank of Canada, McKinsey, Saudi Telecom (STC), ASML, and Accenture as enterprise partners and customers.