Reports and datasets
Original numbers, computed from data TokenZero holds and republished with their method. Download, cite, link back.
5 datasets as of 2026-10-26: a pricing index of 62 models, caching economics for 57 models with a median 90% cached-input discount, the gateway overhead benchmark at 1.52 ms p50 added, the audit corpus scores and the model retirement calendar, each downloadable as JSON and CSV.
· Source: TokenZero pricing table, scripts/bench result files, public audit ledger
At a glance
- Datasets
- 5
- Models in the pricing index
- 62
- Median cache discount
- 90% (computed)
- Gateway overhead p50
- 1.52 ms (computed)
- Formats
- HTML, Markdown twin, JSON, CSV
- Licence
- CC BY 4.0 with attribution
- Refresh
- computed on every request from the live data
- Version
- TokenZero 0.2.5
The datasets
| Report | Rows | As of |
|---|---|---|
| LLM API pricing index, October 2026 | 62 | 2026-10-26 |
| Prompt caching economics across providers, 2026 | 57 | 2026-10-26 |
| Gateway overhead benchmark, October 2026 | 9 | 2026-10-25 |
| LLM token-waste audit corpus and public audit statistics, 2026 | 13 | 2026-10-26 |
| LLM model retirement calendar, 2026 | 17 | 2026-10-26 |
All reports
LLM API pricing index, October 2026Prompt caching economics across providers, 2026Gateway overhead benchmark, October 2026LLM token-waste audit corpus and public audit statistics, 2026LLM model retirement calendar, 2026
Questions people ask
Can I republish these numbers?
Yes, under CC BY 4.0 with a link to the report page; the JSON summary carries every figure quoted in the text and the CSV carries the rows.
How fresh are they?
They are computed on every request from the pricing table (drift-checked daily), the benchmark files produced by the published scripts and the public audit ledger; the as-of date is on each page.
Why no third-party benchmark scores?
TokenZero publishes only what it measures or prices. Third-party quality scores will appear with their source and date once a verified importer ships.