Cohere is a Canadian artificial intelligence company that develops large language models (LLMs) — the underlying software that powers AI systems capable of reading, writing, and analyzing text — built specifically for enterprise customers. Unlike consumer-facing AI products, Cohere's tools are designed to be deployed inside businesses for tasks such as document search, customer support automation, and data summarization. The company is headquartered in Toronto, with offices in New York, London, and San Francisco.
Cohere was founded in 2019 by Aidan Gomez, Nick Frosst, and Ivan Zhang. Gomez and Frosst previously worked at Google $GOOGL Brain, Google's AI research division, and Gomez was a co-author of the landmark 2017 research paper "Attention Is All You Need," which introduced the transformer architecture that underpins most modern LLMs, including those built by OpenAI and Google.
The company's core products include Command, a model built for text generation and business task completion, and Embed, which converts text into numerical representations used for search and classification. Cohere also offers Rerank, a tool that improves the relevance of search results. A central part of Cohere's strategy is retrieval-augmented generation (RAG) — a technique that allows AI models to pull information from a company's own documents and databases rather than relying solely on what the model learned during training, which reduces the risk of the model producing inaccurate or fabricated answers.
Cohere positions itself as a more flexible and privacy-conscious alternative to larger AI providers by allowing its models to run on a company's own cloud infrastructure or on-premises servers, rather than only through Cohere's own platform. This approach is intended to appeal to industries with strict data security requirements, such as finance, healthcare, and government.
