Cheaper Gemini API: what the price cut actually costs
Gemini Flash is priced aggressively already, which makes it a strange place to hunt for a discount. The honest case for a relay here is not the per-token price - it is running Gemini next to your other models without a second SDK, a second key and a second billing relationship.
Our Gemini prices against Google's list
| Model | Official (in/out per 1M) | Tapix AI (in/out per 1M) | Below list |
|---|---|---|---|
gemini-3.6-flash | $0.75 / $3.75 | $0.525 / $2.63 | 30% |
gemini-3.5-flash | $1.5 / $9 | $1.05 / $6.3 | 30% |
gemini-3.7-flash | $0.75 / $3.75 | $0.525 / $2.63 | 30% |
gemini-3.1-pro-preview | $2 / $12 | $1.4 / $8.4 | 30% |
gemini-3.8-flash | $0.825 / $4.13 | $0.5775 / $2.89 | 30% |
What you keep
- OpenAI-style /v1/chat/completions, streaming, tool calls and response_format.
- One key and one prepaid balance shared with Claude, GPT, DeepSeek and others.
- Per-token billing with no monthly seat, starting at a $10 top-up and $1 of free credit.
What you lose
Google-native capabilities: Search grounding, Vertex AI integrations, batch inference, the Google SDK's own features. If your product depends on those, use Google directly - a relay cannot fake them. We also do not expose Gemini's thinking text on our Gemini route today, and we say so on the model listing instead of advertising a feature we do not return.
Worked example
A summarisation service pushing 40M input tokens and 4M output tokens a month on a Flash-class model: at list prices that is a few hundred dollars; the same traffic through a relay that is 30% below list saves roughly the price of a small VPS. That is the whole pitch - not a new business model, just a smaller line item.
Prices above come from our live catalogue. The pricing page is always the source of truth.
Get $1 free creditSee all models