> For the complete documentation index, see [llms.txt](https://nfa.gitbook.io/nfa-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://nfa.gitbook.io/nfa-docs/the-forecasting-engine/cost-and-reproducibility.md). # Cost and reproducibility ## Cost and reproducibility A forecasting engine that you pay for, and trust, has to be honest about two things: what a run cost, and what it actually did. NFA is built so both are verifiable rather than taken on faith. ### Per-purpose model routing Different parts of a run have different needs. The ruling deserves a strong model; lighter steps can run on cheaper ones. NFA routes each step to an appropriate model rather than using one model for everything, which keeps quality where it matters and cost down where it does not. The verification step deliberately uses a separate, independent model from the judge, so a forecast is never checked by the same reasoning that produced it. ### Honest cost accounting Every model call is metered in USD as it happens. Those totals are reconciled against the provider's own billing records to a small tolerance, so the cost attributed to a run is the real cost, not an estimate. Repeated context within a run is cached to cut both latency and spend. {% hint style="info" %} This is why run pricing can be tied to actual usage rather than a flat guess. See [Credits](/nfa-docs/the-economy/credits.md) for how compute cost becomes a credit price. {% endhint %} ### Reproducibility and audit Every run records what it did: which models ran, which options were active, what data anchored it, and what the self-verification concluded. Combined with deterministic data routing, where the same market always resolves to the same source and value, this makes a forecast reconstructable after the fact. You are never asked to trust a number with no provenance; the provenance is part of the record. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://nfa.gitbook.io/nfa-docs/the-forecasting-engine/cost-and-reproducibility.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.