# DeepSeek vs Cohere

> DeepSeek offers MIT-licensed models with 1M context at very low first-party prices, but hosts data in China. Cohere costs more and targets enterprises that need private deployment.

Canonical: https://www.subconscious.dev/compare/deepseek-vs-cohere · By The Subconscious Team · Updated September 30, 2026

## How they compare

On model strength and price, DeepSeek leads. V4 Pro costs $1.32 in and $3.96 out at peak and V4.1 Flash $0.30 in and $1.20 out, with every off-peak hour at half price and cache hits at a few cents per million or less. Both carry a 1M window with 384K max output. Cohere's Command A lists at $2.50 in and $10 out with 256K context, Command A+ runs 128K, and Cohere acknowledges Command A+ trails the latest DeepSeek models on agentic coding. Both publish open weights, MIT for DeepSeek and Apache 2.0 for Command A+.

Data location settles it for many buyers. DeepSeek's hosted API stores data in China, a hard stop for many enterprises, and frequent retirements and repricing keep cost models moving. Cohere sells to banks and governments, runs on Bedrock, Azure AI Foundry and OCI, and deploys into any VPC or fully on-prem with fine-tuning. It also brings Embed 4 and Rerank 4 for retrieval. Teams comfortable self-hosting or using a third-party host can get DeepSeek's weights without the China issue. Teams that want a vendor-supported private stack will find Cohere simpler.

## What each one does

### DeepSeek

DeepSeek is the Chinese lab whose open-weight models reset price expectations for the whole market. Its API now serves two models, both with 1M context and 384K max output. V4.1 Flash shipped September 10, 2026 with built-in image understanding at $0.30 in and $1.20 out at peak. V4 Pro, generally available since August 13, costs $1.32 in and $3.96 out at peak. Cache hits cost a few cents per million or less, and the weights ship on Hugging Face under an MIT license.

### 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.

## Which is best, and when

### Choose DeepSeek for

- Cost-sensitive agents scheduled off-peak
- 1M context with 384K output
- Self-hosting MIT-licensed frontier weights

### Choose Cohere for

- Regulated buyers barred from China-hosted APIs
- Vendor-supported on-prem deployment
- Reranking and embeddings for enterprise search

## At a glance

| Attribute | DeepSeek | Cohere |
|---|---|---|
| Model access | Open weights (MIT) | Closed, plus open Command A+ |
| Flagship models | DeepSeek V4.1 Flash, V4 Pro | Command A+, Command A, Embed 4, Rerank 4 |
| Speed | ~35 tok/s on V4 Pro | 375 tok/s on Command A+ W4A4, per Cohere |
| Price | Off-peak hours at half price | $0.0375–$2.50 in, $0.15–$10 out per 1M |
| Customization | Open weights to fine-tune | Enterprise fine-tuning, incl. private |
| Deployment | First-party API, Hugging Face weights | API, Bedrock, Azure, OCI, VPC, on-prem |
| Long context | 1M, 384K max output | 256K on Command A; 128K on A+ |

## FAQ

### What is the difference between DeepSeek and Cohere?

DeepSeek offers MIT-licensed models with 1M context at very low first-party prices, but hosts data in China. Cohere costs more and targets enterprises that need private deployment.

### When should I choose DeepSeek over Cohere?

Cost-sensitive agents scheduled off-peak; 1M context with 384K output; Self-hosting MIT-licensed frontier weights.

### When should I choose Cohere over DeepSeek?

Regulated buyers barred from China-hosted APIs; Vendor-supported on-prem deployment; Reranking and embeddings for enterprise search.

### Is DeepSeek or Cohere cheaper?

DeepSeek: Off-peak hours at half price. Cohere: $0.0375–$2.50 in, $0.15–$10 out per 1M. The cheaper choice depends on the model and workload.

### Which has more context, DeepSeek or Cohere?

DeepSeek: 1M, 384K max output. Cohere: 256K on Command A; 128K on A+.

## Running long-horizon agents?

If your agents run past 200K tokens, compare both against Subconscious: [Subconscious vs DeepSeek](https://www.subconscious.dev/compare/subconscious-vs-deepseek.md), [Subconscious vs Cohere](https://www.subconscious.dev/compare/subconscious-vs-cohere.md).

Full profiles: [DeepSeek](https://www.subconscious.dev/providers/deepseek.md), [Cohere](https://www.subconscious.dev/providers/cohere.md).