Methodology

How we keep the numbers honest.

TokenTally’s estimates are only as good as the data underneath. Here’s exactly how we ingest provider pricing, normalize it, and audit changes before the calculator sees them.

Dataset snapshot

43 models tracked

Last synced

3/13/2026

Our dataset lives in a private, version-controlled workspace so only vetted changes reach production.

Every revision includes citation URLs and capture times, giving us a full audit trail without exposing internal paths.

Workflow

Collect

We pull pricing directly from provider pages (OpenAI, Anthropic, Google, Mistral, Meta/Fireworks).

Normalize

All entries are converted to USD per million tokens and tagged with context windows + latency classes.

Review

Each proposed change goes through a human-reviewed diff with source citations before it hits production.

Publish

Our curated dataset drives the calculator UI, preset scenarios, and the comparison table instantly.

Normalization rules

All costs are stored as USD per million tokens for both input and output usage.
Context windows are recorded in raw tokens; we warn if a scenario exceeds the selected model.
Latency classes (economy/standard/premium) are qualitative flags to help filter comparisons quickly.

Update cadence

Providers rarely update prices without notice, but we still run a manual review every Friday. If a model is deprecated or a new context window lands, it gets logged before going live.

Quality guardrails

Every update includes a citation URL and the date we captured it.
Automated regression tests cover the calculator math so refactors don’t silently change outputs.
Future milestone: wire scrapers + screenshots so reviewers can sanity-check diffs faster.