Our take
Google AI is the first-party platform for Gemini and Gemma models, offering a tightly curated 16-model catalogue with a clear price ladder from budget flash-lite variants up to premium tiers. It is the direct route to newest Gemini preview releases and lightweight inference at the lowest entry price in its own family.
Use this for first-party access to new Gemini preview releases, or for lightweight inference where the cheapest tier in the family suffices. Pick it for pricing straight from the model originator. Skip it if you need verified SOC 2, a confirmed EU endpoint, or a zero-retention guarantee; all are unverified in our data.
- Lowest entry price in its own catalogue — Gemini 2.5 Flash Lite is 2.5× cheaper on input than the next tier up.
- Broad coverage of Gemini Flash generations: 12 distinct offers spanning 2.5 through 3.8, plus preview channels.
- Direct first-party access to preview models before stable release.
- Compliance posture largely undocumented: SOC 2, EU endpoint, zero-retention, and prompt training policy all unverified in our data.
- Throughput unknown across all offers — no tokens-per-second measurements held.
- Premium tier escalates steeply from entry level: Gemini 3.5 Flash is 15× more on input and 22.5× more on output than the cheapest tier.
Point your tools here
We have not recorded what a router needs for Google AI yet — no base URL and no docs link on file. That is our gap, not a sign Google AI has no API; its own documentation is the place to look until we close it.