# Cohere

> Enterprise-first models for RAG, search and agents, built to run in a private cloud or on-prem.

Canonical: https://www.subconscious.dev/providers/cohere · By The Subconscious Team · Updated September 30, 2026

- Founded: 2019
- Example models: Command A+, Command A, Embed 4, Rerank 4
- Website: https://cohere.com

## Overview

Cohere is a Toronto-based lab that sells models and platforms to banks, governments and large enterprises rather than consumers. Its generative line is the Command family. Command A+, released May 20, 2026, is a 218B-parameter mixture-of-experts model with 25B active, published under Apache 2.0 with a 128K context window, and it combines reasoning, vision, translation and tool use in one set of weights. Command A has a 256K window and lists at $2.50 in and $10 out per million tokens, while Command R7B costs $0.0375 in. June 2026 added North Mini Code, a 30B Apache 2.0 coding model, and the lineup also includes Aya multilingual models and Transcribe for speech.

Retrieval is where Cohere is strongest. Embed 4 handles text, images and PDFs with a 128K context, and Rerank 4 comes in Pro and Fast versions priced per search of up to 100 documents. The models run on Cohere's API and on Amazon Bedrock, SageMaker, Azure AI Foundry and Oracle OCI, though newer models reach each cloud at different times. Model Vault offers dedicated managed instances from $4 an hour. For stricter setups Cohere supports private deployment in any VPC or fully on-prem, including fine-tuning inside that environment, and North is its platform for building internal agents on top. Cohere also agreed in 2026 to combine with Germany's Aleph Alpha.

## Upsides

- Private VPC and on-prem deployment, including fine-tuning, is a core product rather than an add-on.
- Embed and Rerank are a mature, cheap retrieval stack that works alongside any generator.
- Command A+ is open under Apache 2.0 and runs on two H100s or one B200 in 4-bit form.

## Core use cases

- Regulated enterprises that need RAG and agents inside their own network.
- Adding reranking or multimodal embeddings to an existing search system.
- Multilingual assistants and translation across dozens of languages.

## Downsides

- Per-token prices for Command A+, Reasoning, Vision and Translate are not published, so production use often starts with sales.
- Command A+ trails the latest DeepSeek, GLM and MiniMax models on agentic coding and broad intelligence indexes.

## At a glance

| Attribute | Value |
|---|---|
| Model access | Closed, plus open Command A+ |
| Flagship models | Command A+, Command A, Embed 4, Rerank 4 |
| Speed | 375 tok/s on Command A+ W4A4, per Cohere |
| Price | $0.0375–$2.50 in, $0.15–$10 out per 1M |
| Customization | Enterprise fine-tuning, incl. private |
| Deployment | API, Bedrock, Azure, OCI, VPC, on-prem |
| Long context | 256K on Command A; 128K on A+ |

## FAQ

### What is Cohere?

Cohere is a Toronto-based lab that sells models and platforms to banks, governments and large enterprises rather than consumers. Its generative line is the Command family. Command A+, released May 20, 2026, is a 218B-parameter mixture-of-experts model with 25B active, published under Apache 2.0 with a 128K context window, and it combines reasoning, vision, translation and tool use in one set of weights. Command A has a 256K window and lists at $2.50 in and $10 out per million tokens, while Command R7B costs $0.0375 in. June 2026 added North Mini Code, a 30B Apache 2.0 coding model, and the lineup also includes Aya multilingual models and Transcribe for speech.

### What is Cohere best for?

Regulated enterprises that need RAG and agents inside their own network; Adding reranking or multimodal embeddings to an existing search system; Multilingual assistants and translation across dozens of languages.

### How much does Cohere cost?

Cohere pricing at a glance: $0.0375–$2.50 in, $0.15–$10 out per 1M. Rates change often, so check Cohere's pricing page before committing.

### How much context does Cohere support?

Cohere's long-context support: 256K on Command A; 128K on A+.

### What are the downsides of Cohere?

Per-token prices for Command A+, Reasoning, Vision and Translate are not published, so production use often starts with sales; Command A+ trails the latest DeepSeek, GLM and MiniMax models on agentic coding and broad intelligence indexes.

### What are the best alternatives to Cohere?

Common alternatives include Subconscious, OpenAI, Anthropic, Google Vertex AI, Amazon Bedrock. Each has a head-to-head comparison with Cohere on this site.

## Comparisons

- [Subconscious vs Cohere](https://www.subconscious.dev/compare/subconscious-vs-cohere.md)
- [OpenAI vs Cohere](https://www.subconscious.dev/compare/openai-vs-cohere.md)
- [Anthropic vs Cohere](https://www.subconscious.dev/compare/anthropic-vs-cohere.md)
- [Google Vertex AI vs Cohere](https://www.subconscious.dev/compare/google-vertex-vs-cohere.md)
- [Amazon Bedrock vs Cohere](https://www.subconscious.dev/compare/aws-bedrock-vs-cohere.md)
- [Together AI vs Cohere](https://www.subconscious.dev/compare/together-ai-vs-cohere.md)
- [Fireworks AI vs Cohere](https://www.subconscious.dev/compare/fireworks-vs-cohere.md)
- [Baseten vs Cohere](https://www.subconscious.dev/compare/baseten-vs-cohere.md)
- [Groq vs Cohere](https://www.subconscious.dev/compare/groq-vs-cohere.md)
- [Cerebras vs Cohere](https://www.subconscious.dev/compare/cerebras-vs-cohere.md)
- [DeepInfra vs Cohere](https://www.subconscious.dev/compare/deepinfra-vs-cohere.md)
- [Hugging Face Inference Providers vs Cohere](https://www.subconscious.dev/compare/hugging-face-vs-cohere.md)
- [Modal vs Cohere](https://www.subconscious.dev/compare/modal-vs-cohere.md)
- [Cloudflare Workers AI vs Cohere](https://www.subconscious.dev/compare/cloudflare-workers-ai-vs-cohere.md)
- [xAI vs Cohere](https://www.subconscious.dev/compare/xai-vs-cohere.md)
- [Mistral AI vs Cohere](https://www.subconscious.dev/compare/mistral-ai-vs-cohere.md)
- [DeepSeek vs Cohere](https://www.subconscious.dev/compare/deepseek-vs-cohere.md)
- [Moonshot AI vs Cohere](https://www.subconscious.dev/compare/moonshot-ai-vs-cohere.md)
- [Z.ai vs Cohere](https://www.subconscious.dev/compare/z-ai-vs-cohere.md)
- [Alibaba Cloud vs Cohere](https://www.subconscious.dev/compare/alibaba-cloud-vs-cohere.md)
- [Meta vs Cohere](https://www.subconscious.dev/compare/meta-vs-cohere.md)
- [Cohere vs SambaNova](https://www.subconscious.dev/compare/cohere-vs-sambanova.md)
- [Cohere vs Nebius](https://www.subconscious.dev/compare/cohere-vs-nebius.md)
- [Cohere vs Crusoe](https://www.subconscious.dev/compare/cohere-vs-crusoe.md)
- [Cohere vs fal](https://www.subconscious.dev/compare/cohere-vs-fal.md)
- [Cohere vs Novita AI](https://www.subconscious.dev/compare/cohere-vs-novita-ai.md)
- [Cohere vs Venice](https://www.subconscious.dev/compare/cohere-vs-venice.md)
- [Cohere vs Parasail](https://www.subconscious.dev/compare/cohere-vs-parasail.md)
- [Cohere vs Inference.net](https://www.subconscious.dev/compare/cohere-vs-inference-net.md)
- [Cohere vs GMI Cloud](https://www.subconscious.dev/compare/cohere-vs-gmi-cloud.md)
- [Cohere vs Thinking Machines](https://www.subconscious.dev/compare/cohere-vs-thinking-machines.md)
- [Cohere vs Sail Research](https://www.subconscious.dev/compare/cohere-vs-sail-research.md)
- [Cohere vs Morph](https://www.subconscious.dev/compare/cohere-vs-morph.md)
- [Cohere vs Relace](https://www.subconscious.dev/compare/cohere-vs-relace.md)
- [Cohere vs TypeSafe AI](https://www.subconscious.dev/compare/cohere-vs-typesafe-ai.md)
- [Cohere vs StepFun](https://www.subconscious.dev/compare/cohere-vs-stepfun.md)
- [Cohere vs Runware](https://www.subconscious.dev/compare/cohere-vs-runware.md)
- [Cohere vs StreamLake](https://www.subconscious.dev/compare/cohere-vs-streamlake.md)
- [Cohere vs Wafer](https://www.subconscious.dev/compare/cohere-vs-wafer.md)
- [Cohere vs RunInfra](https://www.subconscious.dev/compare/cohere-vs-runinfra.md)
- [Cohere vs Particle.AI](https://www.subconscious.dev/compare/cohere-vs-particle-ai.md)

## Sources

- [Cohere models overview](https://docs.cohere.com/docs/models)
- [Cohere pricing](https://cohere.com/pricing)
- [Command A+ launch, VentureBeat](https://venturebeat.com/technology/cohere-cracks-lossless-quantization-and-native-citations-with-first-full-apache-2-0-licensed-open-model-command-a)
- [Cohere pricing 2026, eesel AI](https://www.eesel.ai/blog/cohere-ai-pricing)

Pricing and model lineups change often; figures are a snapshot.