Pricing

Usage costs depend on whether you’re accessing AI21 models through the AI21 platform (REST endpoints, the SDK, or the AI21 playground) or through a third-party host (a cloud platform), which endpoints you use, and whether you have a custom payment plan with AI21.

Pricing is typically calculated either by call or by token:

• Cost per call means billing per API request. An API request is required for each call to a model, whether by API, SDK, or using the playground.
• Cost per token, where input and output tokens may be priced differently. Token usage counts are given in the usage field in API responses.

AI21 platform usage

Usage costs for the AI21 Platform (REST endpoints, the SDK, or the AI21 playground) are given here.

Free trial usage

New accounts are given a $10 credit good for three months on the AI21 platform. You can use your credit for all usage of the APIs, the SDK, and the playground. Once your trial usage is exceeded or expires, you must provide valid billing information in your account to continue using the AI21 platform.

Subsequent usage

Once you have exceeded your free trial period or credit, you must provide valid billing information to continue using AI21 services. If you have large usage expectations, consider signing up for a customized plan.

Usage is charged monthly from account creation date. Per-token costs are per thousand (K) or per million (M), with no rounding. So 1,500 tokens at $0.10/K tokens cost (1,500 / 1,000) * $0.10 = $0.15

See your AI21 Platform usage and billing

• See your token counts and costs: Navigate to Account > Model usage
• See or change your billing plan: Navigate to Account > Billing & Plans

Cloud usage

AI21 models are hosted on several cloud providers, including AWS SageMaker and Bedrock, and Microsoft Azure. Cloud usage costs are set by the cloud provider. For example, SageMaker charges rates based on usage time and instance type, whereas Bedrock charges a per-token fee.