Gemini provider

Setting up and using Google Gemini (3.6 Flash, 3.5 family, 3.1 Pro, 2.5 family) in Cerevisor.

Setup

You need an API key from aistudio.google.com/apikey.

  1. Settings → Providers → + Add provider → Gemini.
  2. Paste your API key into the API key field.
  3. Click Test connection. Cerevisor sends one tiny request and checks that a real answer comes back. It waits up to 30 seconds.
  4. Click Save.

The key is stored in your OS keychain. It never lives in plain text on disk or in any .cerevisor file.

Missing part of the app: if Gemini calls fail because this copy of Cerevisor is missing the part that talks to Gemini, reinstall Cerevisor from cerevisor.com — your workflows and settings are kept. You are never asked to repair an installed app yourself.

Models

Cerevisor asks Google which models your account can use and shows them everywhere a model picker appears. If Google can't be reached, Cerevisor shows a list it already has, so the picker is never empty.

The most-used Gemini models you'll see:

Model Best for Notes
gemini-3.8-flash General-purpose agents — coding, agentic planning, multimodal. Released September 2, 2026. The default if no model is set. Same half-price introductory rates as 3.7 Flash through December 31, 2026 (they double on January 1, 2027). 1M context, 64K max output.
gemini-3.7-flash General-purpose agents — coding, agentic planning, multimodal. Released August 2026. Half-price introductory rates through December 31, 2026 (they double to 3.6-flash rates on January 1, 2027). 1M context, 64K max output.
gemini-3.6-flash General-purpose agents — coding, agentic planning, multimodal. Released July 2026. Cheaper output than 3.5 Flash at the same input rate.
gemini-3.5-flash-lite High-volume, low-latency subagents. Released July 2026. The cost-effective pick for reviewers, classifiers, and bulk transforms.
gemini-3.5-flash Sustained frontier performance on long agentic and coding tasks. Highest output rate of the Flash tier.
gemini-3.1-pro-preview Highest quality, deepest reasoning. Per-million pricing doubles above 200K input tokens. Use for senior agents, large-context synthesis.
gemini-3.1-flash-lite Fast and cheap. Strong workhorse for reviewers, classifiers, and high-volume transforms.
gemini-3-flash-preview Balanced quality at moderate cost.
gemini-2.5-pro Still current. Cheaper than 3.1-pro for similar quality on many tasks. Also has the >200K tier pricing bump.
gemini-2.5-flash / 2.5-flash-lite Lower-cost alternatives if you don't need 3.x reasoning.

Gemini 3.5 Flash Cyber (announced alongside 3.6 Flash) and Gemini 3.8 Flash Cyber (announced alongside 3.8 Flash) are not usable from Cerevisor: Google serves them only to vetted defenders through limited-access programs (CodeMender, the Fairwind Program), not the public Gemini API.

If you type a custom model id (e.g. a dated variant like gemini-3.1-pro-preview-2026-04), Cerevisor prices it like the model it is a variant of. For a model with no known price, Cerevisor warns you that your spending limit can't see its cost. See Keeping your models current.

Cost

Cerevisor works out cost from the usage Google reports for each response and the model's price. The status bar shows running cost during a run; the audit log saves the breakdown.

Two Gemini-specific quirks Cerevisor handles automatically:

  • Context-tier pricing. On gemini-3.1-pro-preview and gemini-2.5-pro, input rates DOUBLE and output rates jump roughly 50% when the prompt crosses 200,000 tokens. Cerevisor applies the right rate to each request based on its actual size.
  • Thinking tokens bill as output. Per Google's pricing page, "thinking tokens are included in output pricing." Cerevisor counts thinking as output, so its cost numbers match Google's billing console.

Audio inputs are currently priced at the text rate. This can under-report their cost, never over-report it.

Default model

In the Library entry, you can set a default model for this provider. An agent whose model is set to Auto uses this default. If no default is set, Cerevisor uses gemini-3.8-flash.

Temperature on Gemini 3.x

Google deprecated the temperature, top_p, and top_k sampling parameters in July 2026. On Gemini 3.x models they are ignored, and Google has said future model generations will reject them outright. Cerevisor therefore stops sending temperature to Gemini 3.x and later; it still sends it on 2.5, where it works. Setting an agent's temperature on a 3.x model is harmless but has no effect — steer output with system instructions instead.

Thinking modes

Gemini's reasoning models can think before they answer. The thinking level you set on an agent (Agent Config → Reasoning), from Off to Ultra, sets how much: each step up allows more thinking.

If no level is set, the model decides for itself. Off switches thinking off, which matters on small tasks, where thinking can otherwise use up the whole answer. The top levels are limited to what each model family supports. Higher levels cost more but improve answers on multi-step problems. The setting is ignored on models that don't think.

Tools

Gemini agents use Cerevisor's tools like any other agent. Nothing to set up.

Prompt caching

When the model supports it, Cerevisor prices cached input at Google's lower cached rate instead of the standard input rate. This is fully automatic; you don't need to configure anything.

Common errors

Error What it means Fix
"This copy of Cerevisor is missing the part that talks to Gemini" A part of the app this needs did not load. Reinstall Cerevisor from cerevisor.com — your workflows and settings are kept.
401 / API key invalid The key isn't accepted by Google. Generate a new one at aistudio.google.com/apikey and re-test.
The connection check timed out The test request didn't return within 30 seconds. A test that times out is reported as a failure, never as a pass. Check your network. The Gemini API can be slow during regional incidents.
A model has no price Cerevisor has no price for that model yet. Press Refresh models. See Keeping your models current.

Where to go next

Back to docs