Google Gemini 3 pricing in 2026: consumer plans and API costs

Gaia Banfi
Prezzi di Google Gemini 3 nel 2026 piani per gli utenti e costi delle API
In this article
Want similar results?
Discover how LumenONE can transform your customer management.
Learn More
Back to Blog

In 2026 the pricing of Google Gemini 3 follows two separate tracks. On one side there are consumer plans, on the other the API costs aimed at those who build applications.

The distinction matters because the two pricing models address different needs. Consumer plans involve a monthly fee, while the API is billed based on token consumption.

The Gemini 3 model family

The Gemini 3 generation, released in 2026, includes four models with different balances between speed, reasoning depth and cost. Gemini 3.5 Flash is geared toward agentic tasks and coding.

Gemini 3.1 Pro is the flagship reasoning model, with a context window of one million tokens. Gemini 3.1 Flash-Lite is the budget version for high volumes, while Gemini 3.1 Deep Think is dedicated to extended reasoning.

The consumer plans

Consumer plans are available through Google One. The free plan costs nothing and includes Gemini 3.5 Flash with limited access to Gemini 3.1 Pro, along with tools such as Deep Research and Canvas.

The AI Plus plan costs 7.99 dollars per month and offers twice the usage limits compared with the free version, with 200 GB of storage. The AI Pro plan costs 19.99 dollars per month, with four times the limits and 5 TB of storage.

The AI Ultra plan ranges from 99.99 to 199.99 dollars per month. It unlocks Gemini 3.1 Deep Think and much higher usage limits, along with Cloud credits and other bundled services.

The pay-as-you-go API costs

Those who build applications pay per million tokens, with input and output counted separately. Gemini 3.5 Flash costs 1.50 dollars per million input tokens and 9.00 dollars for output.

Gemini 3.1 Flash-Lite is the cheapest option, at 0.25 dollars for input and 1.50 dollars for output. Gemini 3.1 Pro starts at 2.00 dollars for input and 12.00 dollars for output on prompts up to 200 thousand tokens.

A 50 percent discount is available on most models through batch requests. Context caching also helps reduce costs when the same prompt is sent multiple times.

Additional costs to consider

Model pricing does not cover every expense. Grounding with Google Search is free for 5,000 requests per month on Gemini 3 models, then costs 14 dollars per 1,000 requests.

Image generation with Imagen 4 ranges from 0.02 to 0.06 dollars per image. Video generation with Veo 3.1 costs from 0.10 to 0.40 dollars per second, depending on quality and resolution.

Original article: eesel.ai

Gaia BanfiLumenIA
I help Italian companies understand and adopt artificial intelligence in a concrete, safe, and measurable way.

You might be interested

See all
    Agenti vocali e intelligenza artificiale perché la voce diventa la nuova interfaccia
    • AI & Automation
    • News

    Voice Agents and Artificial Intelligence: Why Voice Is Becoming the New Interface

    Voice is increasingly becoming the command through which work is handed over to artificial intelligence. As long as systems only answered a question, typing remained the most precise method. Now…

    ⏱ 3 minuti di lettura
    Agentic commerce in Italia il 43 dei consumatori delegherebbe gli acquisti a un assistente AI
    • AI & Automation
    • News

    Agentic Commerce in Italy: 43% of Consumers Would Let an AI Assistant Handle Purchases

    Artificial intelligence is entering shopping more and more. One of the most debated scenarios concerns the possibility that an AI assistant could make purchases on behalf of the end user.…

    ⏱ 3 minuti di lettura
    Virus mentali tra agenti IA lo studio di Anthropic ed EPFL sui payload autoreplicanti
    • News

    Mind Viruses Among AI Agents: The Anthropic and EPFL Study on Self-Replicating Payloads

    On 10 August, researchers from Anthropic and the Swiss Federal Institute of Technology in Lausanne published a preprint documenting the spread of self-replicating instructions from one artificial intelligence agent to…

    ⏱ 3 minuti di lettura