Gemini CLI
Google's Coding Agent powered by their Gemini Models.
Setup
There's no setup required for Gemini. It works out of the box with our built-in credits system.
We pass the cost of these models transparently to you with zero markup or extra fees. You can find the pricing for the models in the links below.
Gemini 4 Hydrogen rollout
Gemini 4 Hydrogen adds an everyday coding option with Low, Medium, High and Extra high effort. High is its default. Availability remains off, including for Preview members, pending owner verification.
Standard pricing is $1.50 input, $0.075 cached input and $7.50 output per million tokens, including thinking. Rates stay flat across the context window. Implicit caching starts at 4,096 prompt tokens; explicit caching is unsupported. See Google pricing.
Its 1,048,576-token window is shared by input, thinking and output, with up to 65,536 output tokens per synchronous request. n8v does not implement Google's long-decode continuation. The compatible runtime uses generic model metadata and may compact earlier. Upstream inputs include text, images, audio, video and PDFs; output is text. The selected runtime still determines attachment and tool support.
Google offers free and paid API access, subject to project quota. Own native access needs a Google API key; a compatible endpoint needs its bearer key and a working Google-backed gemini-4-hydrogen route. Function replay and native/compatible file tools have been verified. Other Google tools have no new controls in n8v. Existing Gemini choices remain available.
Gemini 4 Carbon rollout
Gemini 4 Carbon is being prepared for owner verification and remains off, including in Preview. Once enabled for your account, it offers Low, Medium and High reasoning for difficult coding tasks. Argon remains the faster Gemini 4 option; Carbon can use more time and tokens per turn.
Carbon uses the paid Google API. Your compatible endpoint must route gemini-4-carbon to an eligible paid Google key. Standard pricing is $2 input, $0.10 cached input and $10 output per million tokens through November 30, 2026; $4, $0.20 and $20 from December 1. Thinking counts as output. There is no listed long-context surcharge, and caching is implicit only. See Google pricing.
The upstream window is 1,048,576 tokens shared between input, thinking and output. Google's default output limits depend on the API mode; its long-decode continuation is not supported by n8v's current CLI path. Compatible runtimes may compact earlier because they lack built-in Carbon metadata. The model accepts text, images, audio, video and PDFs upstream and returns text; n8v's attachment and tool support still depends on the selected runtime.
Available Models
- Gemini 4 Argon, in Low, Medium, and High reasoning. Google's flagship for complex coding: $2 per million input tokens and $10 per million output tokens through November 30, 2026, then $4 and $20.
- Gemini 3.1 Pro
- Gemini 3.5 Flash
- Gemini 3.1 Flash Lite
Gemini 3 Pro and Gemini 2.5 Pro stay selectable from your model preferences.
If you connect your own compatible endpoint, such as CLIProxy, its key must be on a paid Google tier to run Gemini 4 Argon. Google does not offer Argon on the free tier.