Token cost and TCO estimator

Per-token prices come from LiteLLM's open-source pricing file. LiteLLM is maintained by BerriAI, not by the model providers, so always confirm prices on the provider's official pricing page (linked below) before making budget commitments. Enterprise discounts, batch pricing, regional pricing, taxes and cache-write surcharges aren't included.

Assumptions: a 30-day month. Each agent iteration re-sends the system prompt, the user input and all previous outputs. The cache hit ratio applies to all input tokens, and cached tokens are billed at the registry's cache-read price (estimated at 50% of the input price when the registry has none). Provider caching rules (minimum prefix length, time-to-live) differ, as described in each provider's caching documentation. Results are planning estimates, not quotes.

Platform costs beyond tokens (embeddings, vector search, reranking, guardrails, observability, compute, network and storage) are entered by you from quotes or current bills. There is no official price registry for them, so we don't pre-fill them. Euro amounts use the European Central Bank's daily euro reference rate, which the ECB publishes for information only; your provider's billing currency and rate apply to invoices.