AIDP Unit Pricing for Generative AI Models

An AI Data Platform Unit (AIDP Unit) is the unit of measurement used to meter generative AI model consumption in Oracle Analytics Cloud.

For Generative AI models, the rate at which AIDP Units are charged depends on the model and the number of tokens processed. Tokens are metered per 1,000,000 tokens, across input, cached input, and output.

Token consumption is billed separately based on the applicable Oracle Cloud Infrastructure (OCI) Generative AI pricing for the selected model, including the model type, version, and its associated input and output token usage. Token charges are converted into AIDP Units using the equivalent amount in the customer's local currency, based on the OCI Generative AI pricing in effect at the time of the most recent Oracle Analytics Cloud update. Model availability and supported model versions are determined by OCI Generative AI and may change over time. Current OCI Generative AI pricing is published on the Oracle Cloud Infrastructure pricing page and is subject to change. See OCI Price List.

For Generative AI features, you select from the models available in your instance, for example xAI Grok 4.3, OpenAI gpt-5.2, gpt-5.4, gpt-5.5, and gpt-oss-120b. More models may be added over time. Model availability is subject to the OCI region where your Oracle Analytics Cloud instance is located. Oracle reserves the right to change the specific models available in an OCI region. Some features federate a single request across multiple models.

Note:

Cached input tokens are input tokens served from the model's context cache and are charged at a reduced AIDP Unit rate. Models without a cached tier, for example, gpt-oss-120b, charge cached tokens at the standard input rate. Several models are priced in two context bands: Grok 4.3 uses the higher rates above a 200K-token context, and gpt-5.4 and gpt-5.5 use the higher rates for requests above 272K input tokens (the long-context rate applies to the entire request). Generative AI models are charged only for the tokens processed, and there is no idle or baseline charge.