What you see
- Spend. What Opengeni credits charged, plus a list-price estimate for
calls paid by a connected plan (ChatGPT, Claude, SuperGrok) or your own API
key. Those calls aren’t charged by Opengeni, so their amount is marked with
~and labelled as an estimate. - Model calls, tokens and cache hit rate. The cache hit rate is the share of input the provider read back from its prompt cache.
- Tokens and cost by type. Input splits into new input, cache reads and cache writes; output includes reasoning. Each type shows its token count and its share of the list-price cost, so you can see, for example, that cache writes are a small share of tokens but a large share of cost.
- Over time. One chart per hour (today) or per day, as spend, tokens by type or calls. By model (or by the current breakdown) stacks the bars by group.
- Breakdown by. One table you can group by model, provider, paid with, workspace (organization only), project, session, person or schedule. Sort any column. Select a row to filter the whole dashboard to it and look one level down; remove the filter chip to go back. Recent calls lists the latest individual calls.