jevUse a price with a date
On 20 September 2026, the official models page lists $0.042 per million input tokens and no output-token charge. Check that page before budgeting; a gateway can have different billing.
Work through a small example
Suppose you make 100,000 calls at an average of 1,000 input tokens each. That is 100 million input tokens, or $4.20 at the listed rate. This is an arithmetic example, not measured usage from our labs.
Measure the whole input
Your input includes the state and the questions, including their category descriptions. Estimate from recorded usage rather than counting only the customer’s message. Include retries and any second-stage requests in the call count.
Check what happens after the decision
If every ticket still goes to a text model for a reply, that bill remains. Also count review work: a cheap classification that sends the wrong tickets to the wrong team can cost more than it saves.
Request example
estimated_input_cost = calls * mean_input_tokens / 1_000_000 * 0.042 # Example: 100,000 calls × 1,000 tokens → $4.20 input-only