Cohere

Cohere was founded by Aidan Gomez (co-author of the 'Attention Is All You Need' transformer paper), Nick Frosst, and Ivan Zhang, all University of Toronto alumni, with a focus on enterprise NLP.

Cohere launched Cohere Labs (formerly Cohere For AI), a nonprofit research lab led by Sara Hooker (former Google Brain), contributing open-source fundamental machine learning research.

Cohere launched its Embed API for multilingual semantic search and a Retrieval-Augmented Generation (RAG) platform enabling enterprises to ground LLM outputs in their proprietary data. The platform established Cohere as the leading AI provider for enterprise RAG use cases, differentiating from general-purpose chat-focused competitors.

Cohere became one of fifteen tech companies to sign the voluntary White House commitment on measures for testing, reporting, and research on AI risks, establishing its responsible AI credentials.

Cohere released Command R and Command R+, models specifically optimized for enterprise RAG workloads with tools use, multi-step tool orchestration, and citation capabilities. The models set new benchmarks for retrieval-augmented generation and became the default choice for enterprises seeking accurate, citation-grounded AI responses.

Cohere released Command R+, a model specifically optimized for retrieval-augmented generation (RAG) and enterprise tool use, positioning it as the leading enterprise AI model for production business applications.

Cohere raised $500 million in Series D funding at a $5.5 billion valuation, led by PSP Investments with participation from Cisco, Fujitsu, and AMD. The funding validated Cohere's enterprise-focused AI strategy and provided capital to expand its engineering team and cloud infrastructure for global enterprise deployments.

Cohere launched its North AI platform for enterprise knowledge work alongside multilingual Command models supporting 13 languages. The platform combined AI search, summarization, and writing tools with enterprise-grade security, positioning Cohere as a comprehensive enterprise AI platform provider.

Cohere released Command A, a 111 billion parameter open-weights enterprise model with a 256K token context window, designed for agentic tasks, tool use, RAG, and multilingual performance across 23 languages. The model runs on just two GPUs while matching or exceeding GPT-4o on enterprise benchmarks, making frontier AI accessible for on-premise enterprise deployments.

Cohere released Embed 4, a multimodal embedding model for enterprise search and retrieval over mixed text and image documents. The model features a 128K context window, Matryoshka embedding support, and native binary quantization, achieving the top average score on the MTEB retrieval leaderboard across multiple benchmarks.

Cohere secured an additional $100 million in a second close of its latest funding round, bringing its valuation to approximately $7 billion. New investors included BDC (Business Development Bank of Canada) and Nexxus Capital Management, with the round accompanied by an expanded strategic partnership with AMD for AI infrastructure.

Cohere released Transcribe, a 2 billion parameter open-source automatic speech recognition model under Apache 2.0 license that immediately topped the Hugging Face Open ASR Leaderboard with a 5.42% word error rate. The conformer-based encoder-decoder model outperformed Whisper Large V3, ElevenLabs Scribe v2, and IBM Granite, supporting 14 languages including Arabic, Japanese, and Mandarin.

Cohere acquired German AI firm Aleph Alpha in a deal making the combined company worth approximately $20 billion, backed by $600 million from Schwarz Gruppe, in a bid to challenge US and Chinese AI dominance.

Cohere launched Command A+, its new flagship enterprise AI model, targeting regulated industries including finance, healthcare, manufacturing, and the public sector. The launch reinforced Cohere's enterprise-focused positioning post-Aleph Alpha merger.

Cohere launched North Mini Code, its first agentic coding model, as an open-source 30B-parameter mixture-of-experts model (3B active) under Apache 2.0 license. The model targets developer tooling and sovereign AI deployment, marking Cohere's expansion beyond enterprise RAG into the developer tools market.

Cohere released Parse (parse-v5.0), a 2.3-billion-parameter vision-language model that converts complex enterprise documents including PDFs, presentations, and images into structured Markdown, extracting text, tables, lists, forms, images, captions, and page boundaries. The model has an 8,000-token context window and a 4.6 GB footprint, priced at 1.50 USD per 1,000 pages, and is available via the Cohere API, Microsoft Foundry, and AWS SageMaker.