# Bertrande

> The living registry of digital parameters. What digital services cost, the limits they impose,
> and what changed, when, with the evidence. Every value carries the date it was observed, the
> exact source URL, and a SHA-256 fingerprint of the source content at that moment.

Bertrande is an automated editorial system: programs read published sources and record what they
read. No fact here is produced by a language model. Editorial responsibility rests with a named
person, Paul Roulleau, reachable at corrections@bertrande.com. Reuse is permitted with attribution
and a link.

## How to cite
Cite the service page URL and the observation date shown next to the figure. Figures without a
visible date are not ours.

## Source ranking
Levels 1 to 4 are published. Level 5, meaning third-party aggregators, is used only for
corroboration and is never published on its own. 17465 parameters are
currently held back under that rule.

## Machine formats
Every service page is also available as Markdown at the same address plus `index.md`, and as
`data.json` and `data.csv` in the same folder. The full catalogue with prices is in
`/llms-full.txt`, the price matrix in `/data/prices.json`, and host availability in
`/data/reliability.json`. All of it is CC-BY-4.0 and needs no key.

## Speed, availability, and what our figures are worth
Time to first token and output throughput are in `/data/latency.json`, measured by the Hugging
Face router. Availability per upstream host is in `/data/reliability.json`, measured by
OpenRouter's router. Neither is our own measurement and both say so. We read once a day and keep
every reading, so what we publish is the median of our own dated readings with how many it rests
on, never the latest figure alone: one latency measurement covers a handful of calls. The
computation is ours, the measurements are not. Firms that measure speed cap their public history
at ninety days or seventy-two hours; we keep it. We publish no average of what users report,
because an average of anecdotes is a figure nobody measured.

## Prices announced for a future date
55 prices published by Google for January 1, 2027, each read on the vendor's own price page and stored with a fingerprint of that page. They are not forecasts and not our computation.

- Gemini 3.6 Flash (Google), Cache read price, batch tier: $0.0375 today, $0.075 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache read price, flex tier: $0.0375 today, $0.075 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache read price, priority tier: $0.135 today, $0.270 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache read price: $0.075 today, $0.150 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache storage price, batch tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache storage price, flex tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache storage price, priority tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Cache storage price: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Input price, batch tier: $0.375 today, $0.750 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Input price, flex tier: $0.375 today, $0.750 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Input price, priority tier: $1.35 today, $2.70 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Input price: $0.750 today, $1.50 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Output price, batch tier: $1.875 today, $3.75 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Output price, flex tier: $1.875 today, $3.75 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Output price, priority tier: $6.75 today, $13.50 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.6 Flash (Google), Output price: $3.75 today, $7.50 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache read price, batch tier: $0.0375 today, $0.075 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache read price, flex tier: $0.0375 today, $0.075 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache read price, priority tier: $0.135 today, $0.270 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache read price: $0.075 today, $0.150 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache storage price, batch tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache storage price, flex tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache storage price, priority tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Cache storage price: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Input price, batch tier: $0.375 today, $0.750 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Input price, flex tier: $0.375 today, $0.750 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Input price, priority tier: $1.35 today, $2.70 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Input price: $0.750 today, $1.50 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Output price, batch tier: $1.875 today, $3.75 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Output price, flex tier: $1.875 today, $3.75 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Output price, priority tier: $6.75 today, $13.50 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.7 Flash (Google), Output price: $3.75 today, $7.50 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache read price, batch tier: $0.0375 today, $0.075 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache read price, flex tier: $0.0375 today, $0.075 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache read price, priority tier: $0.135 today, $0.270 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache read price: $0.075 today, $0.150 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache storage price, batch tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache storage price, flex tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache storage price, priority tier: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)
- Gemini 3.8 Flash (Google), Cache storage price: $0.500 today, $1.00 from January 1, 2027 (read 2026-09-18, source https://ai.google.dev/gemini-api/docs/pricing)

## Main entry points
- [Catalogue](https://bertrande.com/en/ai/): all tracked services, 1289 pages
- [The same model, seller by seller](https://bertrande.com/en/compare/): what one model costs from each seller
- [Model families compared](https://bertrande.com/en/vs/): Claude vs GPT, Gemini vs DeepSeek and every other pair, each opening with a short answer computed from dated prices, with price history recovered from captures back to 2023
  that sells it, side by side, every price carrying its own observation date. This is the page
  no single vendor can publish, because it is built by reading several sources independently.
- [Which model for what you are building](https://bertrande.com/en/choose/): say what you are building and see
  the offers that fit, ranked on a stated criterion. No score of ours, no opinion.
- [What changed](https://bertrande.com/en/changes/): every recorded change, newest first
- [Announced](https://bertrande.com/en/announced/): prices a vendor has published for a date that has not arrived
- [Status](https://bertrande.com/en/status/): every source we read, when it last failed and why. Machine copy at `/data/status.json`
- [Which model for what](https://bertrande.com/en/best/): rankings computed from dated figures, criterion stated.
  Each ranking page also serves `answer.json` and `index.md` at its own address, carrying the
  criterion, both counts, the citable sentence and a link to each model's own provenance.
- [Methodology](https://bertrande.com/en/methodology/): how a figure gets published, and what we refuse to do
- [Sources](https://bertrande.com/en/sources/): what we collect and at which level
- [About](https://bertrande.com/en/about/): why there is no invented author here
- [Report an error](https://bertrande.com/en/corrections/): corrections are published as dated changes