Skip to main content
Cohere logo

Cohere

Secure and scalable enterprise and sovereign AI.

chatbots#enterprise-llm#rag#embeddings#open-weights
Free plan Free trial Claimed API Self-hosted Teams
Toolglade’s take

Cohere is a strong fit for enterprises and governments that need secure, private, and compliant AI, especially for RAG, search, and agent workflows where its Embed and Rerank models genuinely shine. Open weights, including a permissive Apache 2.0 flagship, make real self-hosting and sovereignty possible. Trade-offs: flagship generation pricing is premium, there is no image, video, or audio generation, and consumer mindshare lags the largest labs. It is enterprise and sales-led rather than hobbyist-friendly.

About Cohere

Cohere is an enterprise LLM company building the Command generation and agent models, Embed embeddings, and Rerank search models, with a focus on privacy, RAG, and private or on-prem deployment. Its 2026 flagship Command A+ is a Mixture-of-Experts model with vision input and open weights under Apache 2.0. Cohere targets enterprises and governments needing secure, sovereign AI and offers North, an enterprise agent productivity platform.

Cohere focuses on business and sovereign AI rather than consumer chatbots. Its Command family powers generation and agentic tasks, Embed provides multimodal embeddings, and Rerank improves search and RAG relevance. The models emphasize enterprise strengths like retrieval-augmented generation, tool use, multilingual coverage, and the ability to run inside a customer's own cloud or datacenter. As of 2026, the flagship is Command A+, the family's first Mixture-of-Experts model with 218B total and 25B active parameters, added vision input, and open weights under Apache 2.0. The earlier Command A remains a strong dense model with a 256K context window and open weights under a research license, while Command R7B offers a cheap small option. Cohere Labs also publishes open multilingual research models like the Aya family. Cohere sells via API plus private and on-prem deployments, with deep enterprise integrations including SAP, Oracle, and Dell's AI Factory. It is well-capitalized, reaching a roughly $7B valuation in 2025 with Nvidia among backers, and generates most revenue from private deployments in regulated sectors.

TL;DR

Cohere is an enterprise LLM company building the Command generation models, Embed embeddings, and Rerank search models, with a focus on privacy, RAG, multilingual coverage, and private or on-prem deployment. Its 2026 flagship Command A+ is a Mixture-of-Experts model with vision input and open weights under Apache 2.0. It targets enterprises and governments, reached a roughly $7B valuation in 2025 with Nvidia among backers, and earns most revenue from private deployments.

Company overview

Cohere is a Toronto and San Francisco-based enterprise AI company focused on secure, business, and sovereign AI. It builds foundation models sold via API plus private and on-prem deployments, emphasizing data privacy and regulated-industry needs.

The company reports roughly $240M ARR, with about 85% of revenue from private deployments, and remains privately held with no IPO. Cohere Labs is its research arm, publishing open multilingual models.

Product features

Cohere's lineup includes the Command family for generation and agents, Embed 4 for multimodal embeddings, and Rerank 3.5 for search relevance. Command A offers a 256K context and open weights, while Command A+ adds Mixture-of-Experts scaling, vision input, and an Apache 2.0 license.

Strengths include RAG, tool use and agents, multilingual coverage across 23-plus languages, and first-class private, on-prem, and VPC deployment. North is its enterprise agent productivity platform, with deep integrations across SAP, Oracle, and Dell.

Target market

Large enterprises and governments needing secure, private, and compliant AI, including financial services, telecom, public sector, defense, and manufacturing. Cohere is not aimed at consumers.

Buyer personas

End users

Enterprise developers and data scientists building RAG, search, and agent applications on Cohere models.

Buyers

Enterprise and government technology leaders prioritizing security, privacy, and sovereignty.

Key influencers

Compliance, security, and procurement stakeholders in regulated industries.

Ideal customer profile

A large enterprise or government building secure RAG, search, and agent workflows where private deployment, data control, and multilingual support are essential.

Funding & performance

Cohere raised a $500M round in August 2025 at a roughly $6.8B valuation with Nvidia among backers, then extended it in September 2025 with an additional $100M, lifting the valuation to about $7B. Investors include Nvidia and PSP Investments among others. The company is privately held with no IPO.

Pros & cons

Pros

  • Genuinely open weights, including a permissive Apache 2.0 flagship
  • Best-in-class RAG stack combining Embed, Rerank, and Command
  • Strong multilingual coverage across many languages
  • Private, on-prem, and VPC deployment is first-class
  • Large 256K context and vision support on current models
  • Well-capitalized with marquee enterprise partners
  • Free trial tier plus cheap small models for evaluation

Cons

  • Flagship generation pricing is premium, comparable to top rivals
  • No image, video, or audio generation; text, embeddings, and rerank only
  • Consumer and developer mindshare lags the largest labs
  • Command A+ hosted per-token API pricing is not clearly published
  • Enterprise and sales-led motion is less friendly to hobbyists
  • Often trails the very top general-purpose models on raw benchmarks
  • Trial keys are barred from commercial use, requiring paid production keys

Pricing plans

Trial
$0
  • ~1,000 API calls/month
  • All models, non-commercial
  • 20 requests/min
  • Evaluation only
Command A (pay-as-you-go)
$2.50 input / $10.00 output per 1M tokens
  • 256K context
  • Agentic and multilingual
  • Production 500 req/min
  • Open weights (research license)
Embed 4 / Rerank 3.5
Embed ~$0.12/1M tokens; Rerank ~$2.00/1K searches
  • Multimodal embeddings
  • Search reranking
  • RAG optimized
  • Pay-as-you-go
Enterprise / Private Deployment
Custom
  • Private, on-prem, and VPC deployment
  • Command A+ self-host (Apache 2.0)
  • North agent platform
  • Dedicated support

Key features

API
Team collaboration
Self-hosted
Multi-language
Integrations
SAP, Oracle, Dell AI Factory, Amazon Bedrock, Hugging Face
Input types
text, image
Output types
text
Best For
enterprise RAG and search, private and on-prem deployment, multilingual applications, agentic enterprise workflows

Compare key features

View all alternatives →
Feature
Cohere
MiniMax
Reka AI
Pricing
Freemium
Freemium
Freemium
Free plan
Yes
Yes
Yes
Free trial
Yes
No
No
API
Yes
Yes
Yes
Self-hosted
Yes
Yes
Yes
Team support
Yes
No
No

Frequently asked questions

Does Cohere offer open-weight models?+

Yes. Command A and Command R have open weights under a research license, and the 2026 flagship Command A+ is released under the permissive Apache 2.0 license. Cohere Labs also publishes open Aya multilingual models.

Can I self-host Cohere models?+

Yes for open-weight models like Command A+ under Apache 2.0, and private or on-prem deployments are a first-class offering for enterprises, which is a core part of Cohere's business.

Does Cohere generate images, video, or audio?+

No. Cohere focuses on text generation, vision input on current models, multimodal embeddings, and reranking. It does not offer image, video, or audio generation.

Is there a free way to try Cohere?+

Yes. A trial API key allows roughly 1,000 calls per month across endpoints for non-commercial evaluation. Commercial use requires a pay-as-you-go production key.

What is North?+

North is Cohere's enterprise AI-agent productivity platform that automates workflows and syncs data across business apps, and it is integrated into offerings like the Dell AI Factory.

Reviews (0)

Write a review

Pick a rating
Loading reviews…
Compare

Compare Cohere with other AI tools

Side-by-side pages for pricing, features, and best-fit use cases.

All comparisons →

Similar tools you may like