llmproviders.ai
← All providers

Cohere pricing

verified 2026-09-02

Enterprise-focused lab behind the Command family, with strengths in RAG, tool use, and multilingual work.

data residency & retention
from Cohere's own policies · verified 2026-09-02 · some facts not published
Headquarters
🇨🇦 Canada (Other)
Where inference runs
Not disclosed
Cohere does not publicly disclose where its SaaS API platform is hosted; models can alternatively run in customer-controlled environments on OCI, Azure, AWS, and Google Cloud.
Region pinning
Enterprise plans only
No self-serve region pinning for the SaaS API; data residency is achieved via private deployments (VPC/on-prem) or third-party cloud platforms, where Cohere receives no prompts.
Prompt retention
Retained · 30 days
Logged prompts and generations are automatically deleted after 30 days unless required for legal reasons or flagged for policy-violation review; data allowed for training may be retained separately.
Trains on API data
Yes, by default
By default Cohere may use prompts and generations (with personal information filtered) to train its models; customers can opt out any time via dashboard Data Controls.
Zero data retention
Available
ZDR is offered only to enterprise customers who make additional usage commitments; it disables misuse monitoring but usage metadata is still collected.
current models · $ per 1M tokens
ModelInput /1MCached inOutput /1MContext
Command R7B
Command R
$0.04—$0.15128K
Command R
Command R
$0.15—$0.60128K
Cohere totals · 1M in + 1M out
Command R7B
$0.1875
Command R
$0.7500

Cost of 1M input + 1M output tokens. Bar length is square-root scaled so cheap models stay visible next to premium ones.

legacy models · still purchasable
ModelInput /1MCached inOutput /1MContext
Command R+ 08-2024
Command R
$2.50—$10—
Command R+ 04-2024
Command R
$3—$15—
Command R 03-2024
Command R
$0.50—$1.50—
Command
Command
$1—$2—
Command Light
Command
$0.30—$0.60—