Cohere Pricing Plans and Tiers

Cohere's usage-based pricing model charges per API request, creating a cost structure that scales invisibly with production traffic. Unlike seat-based models, usage-based pricing forces teams to estimate token volume, model selection (Command vs. Command-R vs. Command-R+), and inference endpoint overhead before deployment. A team running 100K daily inference requests at $0.0075 per request pays $225/day ($6,750/month)-a cost that's opaque in Cohere's pricing page and only revealed through cost calculator exploration.

The platform offers three model families (Command, Command-R, Embed) with per-family pricing tiers, creating complexity where teams must benchmark inference costs across different model choices. Command-R+ ($0.03 per request) is 4x more expensive than Command ($0.0075 per request), incentivizing cost optimization through model selection rather than feature gates. This is a hidden pricing lever: teams choosing Command-R+ for better reasoning pay 4x the cost without understanding the price-performance tradeoff.

Cohere's 'enterprise' tier bypasses published pricing entirely, targeting 10K+ daily requests or multi-year commitments. This bifurcated pricing (pay-as-you-go vs. custom) creates a jump where scaling from $20K/month usage ($0.0075 x 100K requests) to enterprise requires custom negotiations. Teams deploying Cohere at scale discover that usage-based pricing is unsustainable beyond ~$15K/month, forcing enterprise conversations.