Google Gemini
Connect Google's Gemini API with a key, and how its thinking settings map to the Hub's.
Google's Gemini models through your own Google AI Studio key, billed per token to your Google account.
Set it up#
- Make an API key at https://aistudio.google.com/apikey.
- In Settings → Model Hub → Providers, select the Google Gemini tile and paste the key into API key.
- Select Test.
Models and prices#
| Model | In, per million | Out, per million | For |
|---|---|---|---|
gemini-3.8-flash | $0.75 | $3.75 | Most work |
gemini-3.5-flash-lite | $0.30 | $2.50 | Bulk and quick work |
gemini-3.1-pro-preview | $2.00 | $12.00 | Hard judgment |
gemini-2.5-pro | $1.25 | $10.00 | |
gemini-2.5-flash | $0.30 | $2.50 | |
gemini-embedding-001 | $0.15 | Embeddings |
Prices were checked on 2026-10-06 and are the base rate; Google charges Pro models more for prompts over 200 000 tokens, which the ledger does not model. Google reports no cost per call, so the ledger's cost is the listed price.
Reasoning#
Gemini models think by default. The Reasoning option maps to Gemini's thinkingConfig by generation:
| Model | Off | Low, Medium, High |
|---|---|---|
| Gemini 3 Flash models | thinkingLevel: minimal | the level |
| Gemini 3 Pro | thinkingLevel: low (it goes no lower) | the level |
| Gemini 2.5 Flash | thinkingBudget: 0 | a budget of the Hub's size |
| Gemini 2.5 Pro | thinkingBudget: 128 (its floor) | a budget of the Hub's size |
Role’s setting is Off unless the role asks for a level.
Web research#
Research roles use Gemini's Google Search grounding. Sources come back with the reply.