Gemini field guide: models, plans, and prices

Gemini API pricing: the chat models

These are the text/reasoning models you’d actually pick between. Prices are Gemini Developer API, paid tier, standard rate, per million tokens.

1. a Launch pricing through November 30, 2026. From December 1 Argon is $4.00 input and $20.00 output.
2. b Launch pricing through December 31, 2026. From January 1, 2027 both Flash models are $1.50 input and $7.50 output. If you’re budgeting past the new year, use the second number.

Gemini 4 Argon is the frontier model, in the Gemini app for Pro and Ultra subscribers since October 13. Gemini 3.1 Pro is still the model behind the app’s “Pro” setting and is tiered on the API: above 200k tokens of input it jumps to $4.00 input and $18.00 output. For a new API project there’s little reason to start on 3.1 Pro while the launch prices hold: Argon costs the same on input and less on output, and 3.8 Flash costs a third as much. The two Flash-Lites are the odd pair: the older 3.1 is cheaper than the newer 3.5, so if your task passes on 3.1, keep it.

If you’re starting a project today: Flash-Lite for high-volume work where mistakes are cheap, 3.8 Flash as the everyday default, and Argon when being wrong costs more than the model bill.

Gemini vs Claude vs GPT pricing

The side-by-side table of Gemini, Claude, and GPT prices lives on the field guides hub now, so a price change is one edit instead of three.

Google’s “half to a third the price of frontier models” line only holds against the flagships. Argon at $2.00 / $10.00 matches GPT-6.1 Sol and Claude Sonnet 5.5 to the cent, and costs a fifth of the two $10-input flagships, GPT-6 Astra and Claude Fable 5.1. Its December list price of $4.00 / $20.00 matches Claude Opus 5.5 instead. At the budget end, GPT-6 Luna and Claude Haiku 5.5 are both $0.10 / $0.50, which makes Google’s cheapest model, 3.1 Flash-Lite at $0.25 / $1.50, the expensive cheap model. Pick the cheapest model that passes your own testing, not the one with the best launch slide.

Google AI Pro vs Ultra, and where Plus fits

Prices are US, per month, from Google’s plans page.

Google’s pages disagree about Pro on Plus. The model list still shows it, the plans page does not, and the model-access notice says Plus subscribers lose Pro on a date Google emails them. Treat Plus as Flash-Lite and Flash if you’re deciding today.

The two Ultra tiers are the same plan at different volumes: the $99.99 one is aimed at people who keep hitting Pro’s ceiling, the $199.99 one is the same list at four times the quota, and both now include Project Genie. Plus dropped to $4.99 and doubled its storage to 400 GB, which makes it the cheapest way to get the Flash model and a 128k context window. Most people who pay will still land on Pro: it’s the first tier with Argon, a 1M context window, and Tasks, and the lowest tier that keeps the Pro model.

Gemini agents: Tasks, Antigravity, and Jules

The naming got worse and then better. Gemini is the assistant and the model family. Spark was the agent inside Gemini. This autumn Google renamed Spark to Tasks because, per its own help page, “tasks” was the word people kept using. So the thing that launched as a 24/7 personal AI agent is now a Chat/Task toggle next to the prompt box, the plans page lists it as “Tasks in Gemini”, and the help center has a section titled “Where did Spark go?”

Tasks runs in the background, drives either your local Chrome or a remote browser, and asks before it buys, sends, or submits anything. It needs a Pro or Ultra plan, a personal Google account, and Personal Intelligence switched on. You can have 15 tasks running at once.

Antigravity and Jules are both coding agents but the split is where they run. Antigravity is a desktop app you open and code inside, like VS Code but built around agents instead of around you typing. Your code stays on your machine and you work alongside the agent in real time: it suggests, you accept or reject, you iterate. Jules is the opposite. It’s a cloud service you hand a GitHub task to and then close the tab. It clones your repo into a Google VM, does the work there, and sends back a pull request for you to review. You never touch its environment.

So: Antigravity for hands-on work where you want to watch and steer, Jules for fire-and-forget tasks you’d rather not do yourself (“fix the flaky test, update that dependency”). They bill through the same subscription, so one Pro plan raises your limits on both, and Pro and Ultra can buy AI credits at API prices when the baseline quota runs out. Antigravity 2.0 has a CLI that absorbed the old Gemini CLI, plus an SDK and multi-agent orchestration. Since October 13 it’s also built into the Gemini macOS app, so you can start a coding agent from a chat. If you’ve been using Gemini CLI, Google wants you on the Antigravity CLI now.

Gemini creative models: Omni, Veo, Nano Banana, and Lyria

Gemini Omni Flash is the flagship here and is now generally available on the paid API. Google describes it as a model that creates “anything from any input,” starting with video, with a better grasp of physics like gravity and how fluids move. In practice it’s video generation and editing you drive with plain language: upload a clip, tell it to change the background or add yourself to the scene. The API bills it per output token at 5,792 tokens per second of 720p, which Google’s pricing page puts at about $0.10 a second, the same as Veo 3.1 Fast. In the Gemini app it’s on every paid plan, with limits that scale by tier.

Project Genie is the odd one out and worth explaining since the label “interactive world model” tells you nothing. You type a description of a place (“a hilly landscape of Hobbit homes with little gardens”), and it generates a 3D world you can actually move through in real time, walking or flying, rather than a fixed video you watch. It’s an experimental prototype for Ultra subscribers 18 and over, in 140-plus countries. Think tech demo, not product.

Veo pricing is per second. Veo 3.1 standard is $0.40/sec at 720p or 1080p and $0.60/sec at 4K. Fast is $0.10/sec at 720p, $0.12 at 1080p, and $0.30 at 4K. Lite is $0.05/sec at 720p and $0.08 at 1080p, with no 4K. An 8-second clip runs $3.20 on standard or $0.80 on fast 720p.

Images are all Nano Banana now. Imagen 4 has dropped off the pricing page. Nano Banana 2.1 at about $0.034 per 1K image is the default, and Google claims accurate text rendering, which is the polite way of saying it might finally spell. Nano Banana Pro is four times the price for the hard cases.

The image model naming deserves its own note. The pricing page now lists five things called Nano Banana: Nano Banana 2.1, Nano Banana 2, Nano Banana 2 Lite, Nano Banana Pro, and the original Nano Banana from the 2.5 era. The newer 2.1 is cheaper per image than the 2 it replaces.

Gotchas before you subscribe or build

A few things worth knowing before you subscribe or build.

Personal accounts only. You cannot subscribe to a Google AI plan with a Workspace account. Google’s FAQ is explicit: Workspace customers get a Gemini add-on to their existing subscription instead. If your main account is a Workspace email, you’re either signing up with a personal Gmail or going the add-on route.

Limits are compute-based, not prompt-based. There are no daily prompt counts. A simple text prompt burns far less than a video, Deep Research, or Deep Think prompt. Your allowance refreshes every five hours up to a weekly cap. When you hit the ceiling on a paid plan you drop to Flash-Lite. Pro and Ultra can buy AI credits at API prices to keep going in Flow and Antigravity.

If you don’t pay, you don’t pick the model. Since October 9, free accounts get “Auto”, which routes most prompts to Flash-Lite and only the ones that need it to Flash or Pro. Plus gets Flash-Lite and Flash, and keeps Pro until Google’s announced removal reaches your account. Only Pro and Ultra get Argon, and only they can pick a model for Tasks.

Spark is now Tasks, and still not everywhere. It’s on Pro and Ultra in most countries, but not in the EEA, UK, Switzerland, or Nigeria, and it needs a personal account. Don’t subscribe for the agent before checking the country list.

Region and age limits are everywhere. A lot of features are tagged “US only” (Daily Brief, AI Inbox, Deep Search, Playground), “US and India” (Chrome auto browse), or “18+” (Tasks, Antigravity, Jules, most generative video). The headline list and the list you can actually use in your country are different lists.

The model on the box may not be the model you get. The app’s “Pro” is still Gemini 3.1 Pro, a February model. Argon is a separate pick for Pro and Ultra subscribers. The everyday Flash in the app is 3.6, while the newest Flash on the API is 3.8. If you’re paying for Pro-tier reasoning, check which model is actually serving your requests.

Which Gemini model to use, by job

Two rules cover most of it: pick the cheapest model that passes your own testing, not the one with the best launch slide, and don’t pay for a tier to get one feature that’s still in beta.

Sources, all checked October 21, 2026: Google’s Gemini API pricing, the Google AI plans page, Google’s help pages on Gemini Apps limits and feature availability, Gemini models by plan, changes to model access, Tasks in Gemini, and Google AI Ultra benefits, the Gemini app release notes, the product pages for Antigravity, Jules, and Flow, plus Anthropic’s and OpenAI’s rate cards.