Google Gemini: API key and base URL
Gemini models, official API. Bring your own key; nothing is billed by us.
What the harness already knows
| Provider id | google |
| Base URL | https://generativelanguage.googleapis.com/v1beta |
| Protocol | Google Gemini API |
| API key | GEMINI_API_KEY or GOOGLE_API_KEY |
| Model discovery | No — models come from the catalogue and from what you configure. |
Connecting it
The endpoint above is built in, so there is nothing to configure. Put the key in the environment and the provider appears in the model picker:
export GEMINI_API_KEY="..."Keys are issued by Google Gemini — where to get one.
Models on Google Gemini
The newest models Google Gemini serves that can call tools, from the models.dev snapshot OHarness ships.
| Model id | Context / output | $/M in / cached / out | Reasoning |
|---|---|---|---|
gemini-3.8-flash | 1.05M / 66K | 0.75 / 0.075 / 3.75 | Yes |
gemini-3.7-flash | 1.05M / 66K | 0.75 / 0.075 / 3.75 | Yes |
gemini-flash-latest | 1.05M / 66K | 0.75 / 0.075 / 3.75 | Yes |
gemini-3.5-flash-lite | 1.05M / 66K | 0.3 / 0.03 / 2.5 | Yes |
gemini-3.6-flash | 1.05M / 66K | 0.75 / 0.075 / 3.75 | Yes |
gemini-flash-lite-latest | 1.05M / 66K | 0.3 / 0.03 / 2.5 | Yes |
gemini-3.1-flash-lite-image | 66K / 66K | 0.25 / — / 30 | Yes |
gemini-3.5-flash | 1.05M / 66K | 1.5 / 0.15 / 9 | Yes |
gemini-3.1-flash-lite | 1.05M / 66K | 0.25 / 0.025 / 1.5 | Yes |
deep-research-max-preview-04-2026 | 131K / 66K | 2 / 0.2 / 12 | Yes |
List prices in USD per million tokens, read from models.dev on 2026-09-19. Google Gemini changes its prices and models; its own pricing page is the authority. Start a session on the newest:
oharness -m google/gemini-3.8-flashPointing it somewhere else
A regional host, a gateway, or a server of your own. Overrides in ~/.oharness/config.json are field level, so changing the endpoint does not mean restating the rest of the provider — this is the whole block, shown holding the default so the shape is clear:
{
"providers": {
"google": {
"baseUrl": "https://generativelanguage.googleapis.com/v1beta"
}
}
}What you get either way
The same agent loop, the same tools, the same MCP servers and the same permission rules, whichever provider is underneath — and switching between them mid-session converts nothing, because history is never stored in a vendor’s shape. What an agent harness is covers why that is the hard part, and the providers reference covers the catalogue itself.