Pricing FAQ
This page answers the questions that usually block trust or momentum: where to start, how to read freshness, what the calculator can and cannot claim, and why some rows are labeled differently.
Where should a first-time visitor start?
Start with calculator if you already know token volume and need a monthly spend answer. Start with compare if the workload shape is fixed and you want a side-by-side decision. Start with changes if recent price movement matters most. Start with models if you are still scanning the market.
Why are there both models and providers pages?
Models is the broad market scan. Providers is the ecosystem index. Use models when price sorting matters more than platform boundaries. Use providers when you want one ecosystem summary before drilling into its model detail pages.
When should I use compare instead of the model list?
Use compare once the shortlist is already small and the workload is fixed. The model list is still better for browsing, freshness scanning, and trimming obviously mismatched options.
When should I use changes instead of home or models?
Use changes when the question is “what moved recently?” rather than “what is the current cheapest row?” Change events only appear when a newer stored point differs from the prior value.
What does “Updated” mean?
It is the timestamp of the latest stored pricing snapshot used by the page. It is the primary freshness signal for whether the number is recent enough to inspect further.
What does “Checked” mean?
It is a secondary verification timestamp. When no separate verification pass exists, it can match the latest stored snapshot rather than implying an extra quality step that did not happen.
Why does one model show a verified fallback?
The site is official-first. Verified fallback only appears when the preferred official crawl path is not consistently available, and that limitation stays visible instead of being disguised as an official live crawl.
Why can a model have history but no recent change?
History means stored points exist. A change event only appears when the newer point differs from the previous one. Same-price refreshes do not create fake movement.
Does the calculator match the final provider bill?
No. It is an estimate layer. It normalizes public pricing, then applies the request shape, cache ratio, batch mode, and optional FX conversion shown on the page. Provider-side discounts, credits, taxes, and custom contracts are outside that scope.
Why is display currency different from source currency?
Source currency is the stored provider price basis. Display currency is only the viewing layer. If FX conversion is unavailable, the calculator falls back to the source currency instead of hiding the row or implying false precision.
What do batch counts mean on home and providers pages?
They count models with a listed batch pricing field that the site can normalize right now. That wording is intentionally narrower than blanket provider-wide batch support.
What is the difference between cached pricing and batch discount?
Cached pricing discounts the cached share of input tokens. Batch discount is a separate provider execution mode and should not be treated as the same cost mechanic.
What does the history signal badge mean on model detail?
It compresses stored history depth into a faster trust read. Warming up means more snapshots are needed, low confidence means early movement is visible but still thin, and usable signal means short-term interpretation has enough stored depth to be more dependable.
Why are some models missing from default compare or calculator views?
The default view favors models with live snapshots so the first screen is decision-ready. Models without a usable current snapshot stay separated instead of degrading the main matrix with empty rows.
Does blend replace a real workload?
No. Blend is only a browse-friendly shortcut based on 75% input and 25% output. Once the request shape is known, compare or calculator should replace it.
What if the explanation pages still do not answer my question?
Use glossary for term-level questions, methodology for formulas and data flow, and data sources for trust and source labeling. Those three pages form the full explanation layer around the main product pages.
Best when the question is about the meaning of a field rather than how the site works.
Open glossaryBest when you need to understand how snapshots, deltas, and workload estimates are built.
Read methodologyBest when you need to judge official-first versus verified fallback paths or interpret provider coverage at a higher level.
Review data sources