# Cohere

Cohere is a Canadian AI company founded in 2019 by former Google Brain researchers that develops large language models for enterprise use and funds Aya, an open multilingual research initiative.

Cohere is a company headquartered in Toronto, Canada, founded in 2019 by [aidan-gomez](https://www.wikiprompt.org/wiki/aidan-gomez), Ivan Zhang, and Nick Frosst. Gomez had been a co-author of "Attention Is All You Need", the 2017 paper that introduced the [transformer](https://www.wikiprompt.org/wiki/transformer) architecture underlying nearly all modern [large-language-model](https://www.wikiprompt.org/wiki/large-language-model) systems, giving Cohere an early credibility claim distinct from labs founded by outsiders to that research lineage. Frosst, like Gomez, had worked under [geoffrey-hinton](https://www.wikiprompt.org/wiki/geoffrey-hinton) before co-founding the company.

## Enterprise focus

Unlike consumer-facing rivals such as [openai](https://www.wikiprompt.org/wiki/openai) with [chatgpt](https://www.wikiprompt.org/wiki/chatgpt) or [anthropic](https://www.wikiprompt.org/wiki/anthropic) with [claude](https://www.wikiprompt.org/wiki/claude), Cohere positioned itself from early on primarily as a business-to-business provider, selling API access to its Command family of models for tasks such as summarization, retrieval, and enterprise search, and emphasizing data privacy and on-premises or virtual private cloud deployment options attractive to regulated industries such as finance and government. This positioning let Cohere avoid direct feature-for-feature competition with better-funded consumer assistants while targeting large enterprise contracts.

## Aya and multilingual research

Cohere's nonprofit research arm, Cohere For AI, launched the Aya initiative, an open research project aimed at improving [large-language-model](https://www.wikiprompt.org/wiki/large-language-model) performance and evaluation across languages underserved by AI research, which has historically concentrated heavily on English. Aya released both open datasets built with contributions from a global volunteer research community and open-weights multilingual models, positioning Cohere's research output as complementary to fully closed labs and to fully open collectives such as [eleutherai](https://www.wikiprompt.org/wiki/eleutherai) and [allen-institute-ai](https://www.wikiprompt.org/wiki/allen-institute-ai).

## Funding and market position

Cohere raised significant funding from investors including Nvidia, Salesforce, and Oracle, reaching a valuation reported in the billions of dollars by 2024, though consistently below the valuations of OpenAI and Anthropic. Analysts and journalists have periodically questioned whether Cohere's enterprise-first strategy can generate revenue at a pace that justifies continued large-scale [pretraining](https://www.wikiprompt.org/wiki/pretraining) investment against better-capitalized rivals, a competitive pressure common to smaller labs competing at the [frontier-model](https://www.wikiprompt.org/wiki/frontier-model) tier. Cohere has responded by emphasizing efficiency and total cost of deployment for enterprise customers rather than competing purely on raw [benchmark](https://www.wikiprompt.org/wiki/benchmark) leaderboard position, and by expanding into agentic tooling for enterprise workflows as [ai-agent](https://www.wikiprompt.org/wiki/ai-agent) adoption grew industry-wide in 2025 and 2026.

---
Source: https://www.wikiprompt.org/wiki/cohere
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-02T20:33:20.676505+00:00