CohereAI

CohereAI was founded in 2019 by Aidan Gomez, Ivan Zhang and Nick Frosst. Their headquarters are located in Toronto, Canada and San Francisco, USA. It also has offices in Palo Alto, London, New York and San Francisco.

CohereAI is a company providing a leading artificial intelligence platform. Services they provide include:

Large Language Models (LLMs): Cohere provides industry-leading large language models (LLMs) to help enterprises build powerful, secure applications that can search, understand meaning, and hold conversations as text.

Retrieval Augmented Generation (RAG): Cohere’s RAG toolkit allows LLMs to accurately answer questions and solve tasks using enterprise data as a source of truth.

Semantic Search: Cohere’s Embed model allows enterprises to build powerful search solutions. Embed is the best-performing embedding model in English and over 100 other languages, ensuring relevant results.

Search performance improvements: Rerank greatly improves the relevance of search results from existing lexical and semantic search tools (such as Elasticsearch, OpenSearch or Solr) and can be customized by domain for better performance.

In addition, Cohere also provides a number of other services such as generating unique content (e.g. emails, landing pages, product descriptions, etc.) and assisting enterprises at any stage of their AI journey. Their models are accessible via SaaS APIs, cloud services (e.g. OCI, AWS SageMaker, Bedrock) and private deployments (VPC and on-prem). Cohere takes customer data privacy very seriously, they never use customer data when training base models, and customers have full control over customization and model input/output. Overall, CohereAI is a company dedicated to solving complex machine learning problems.

CohereAI’s official website: https://cohere.com/

CohereAI also created a website called LLM University to disseminate knowledge related to cultivating large models: https://docs.cohere.com/docs/llmu

On March 11, 2024, CohereAI open sourced the Command R model with 35 billion parameters.

On April 4, 2024, CohereAI open sourced the Command R+ model with 104 billion parameters, which was the open source model with the largest parameter scale so far.

CohereAI's open source models Command R and Command R+ are the strongest models in the open source field and have received good reviews in various reviews and feedback. The latest Command R+ model surpassed the early version of GPT-4 in the anonymous voting score of large models in early April 2023. It is also the first time that an open source model has surpassed GPT-4 in the anonymous voting score.