Return Capacity Unit (CU) estimates in Orchestration API responses #1016
Description
Describe the Problem
Currently, the Orchestration API returns token usage (input_tokens, output_tokens), but applications must manually calculate the equivalent Capacity Units (CUs) based on constants that are not available through any official API.
At present, the only publicly available sources for CU pricing details per model are:
• SAP AI Core CU Calculator, which uses hardcoded values from this file: Constants-dbg.js
• OSS Note 3437766 – Availability of Generative AI Models (tables listing models and CU information)
Since CUs are the official way to measure costs in AI Core, developers currently need to manually replicate the pricing logic, which is error-prone and does not automatically reflect updates to the official pricing model.
Propose a Solution
Extend Orchestration API responses to include estimated CU usage for input and output tokens. For example:
Model: gpt-5
{
"input_tokens": 1000000,
"output_tokens": 1000,
"capacity_units_input": 2.0387,
"capacity_units_output": 0.0152
}
Scope:
• Applies only to orchestration calls (chat/completion endpoints).
• CU estimation should be based on the current pricing maintained by SAP, so customers don’t need to hardcode constants.
• Currency conversion is not required in the API response; customers can handle that separately.
Benefits:
• Transparent cost monitoring directly from the API.
• Easier integration of AI consumption metrics into business applications.
• Reduced risk of discrepancies when SAP updates pricing models.
Describe Alternatives
No response
Affected Development Phase
Production
Impact
Inconvenience
Timeline
No response
Additional Context
1 Capacity Unit = €1.04 according to the SAP AI Core Pricing.