Google AI Studio
AI Image Verified- Free sandbox tier Free prototyping Gemini 3.8 Flash
- $0/mo $0/mo $0.75/1M
- Free tier (rate-limited, logged) Free workbench prototyping Gemini 3.8 Flash input tokens
- Paid privacy rule Async Batch discount Gemini 4 Hydrogen
- Zero log −50% off $1.50/1M
- Paid API stops prompt training Async Batch cuts API rates 50% Gemini 4 Hydrogen input (1M ctx)
- Real unit cost
- $0.034/img
- Nano Banana 2.1 Paid API (1K)
Google AI Studio is free to prototype with inside the web workbench and Free API tier, yet moving to the Paid API tier changes two core realities: Google stops using your prompts to improve products, and production rate limits plus Gemini 4 Carbon, Gemini 4 Argon, Nano Banana 2.1 ($0.0336 per 1K image), and Batch API (50% price cut) unlock. Checked live on ai.google.dev/pricing on 2026-11-14.
Pricing summary verified 2026-11-14: Free Tier includes $0 API calls for Gemini 4 Hydrogen, Gemini 3.8 Flash, Gemini 3.8 Flash TTS, and Gemini 2.5 Pro/Flash, while Google retains permission to use Free prompts to improve products. Paid Tier excludes customer data from product improvement and adds prepaid or pay-as-you-go billing with a 50% Batch discount.
Free tier vs Paid tier privacy and quota rules
Many developers start inside Google AI Studio without attaching a Cloud Billing account because Free Tier access costs zero dollars for testing prompts. That convenience comes with strict terms and feature gates that force commercial products onto the Paid Tier before launch.
- Prompts and outputs submitted on the Free Tier can be used by Google reviewers and pipelines to improve Google products, making it unsuitable for client confidentiality, customer support logs, or proprietary datasets.
- Flagship reasoning and media endpoints including Gemini 4 Carbon, Gemini 4 Argon, Nano Banana 2.1, Gemini 3 Pro Image, and Lyria 3.5 Music return permission errors on the Free Tier and require active billing.
- Batch API scheduling (50% token and image discount for 24-hour asynchronous jobs) and explicit Context Caching storage also remain locked until billing is enabled.
Gemini 4, Flash, and media token rates compared
Pay-as-you-go billing counts every processed token, including model thinking tokens and system tool schemas when grounding is enabled. Notice the scheduled rate adjustments taking effect December 1, 2026 for Gemini 4 flagship models and January 1, 2027 for Flash models.
| Model | Free tier | Input price (per 1M tokens) | Output or unit price | Scheduled rate change / note |
|---|---|---|---|---|
| Gemini 4 Carbon / Argon | Not available | $2.00 | $10.00 / 1M tokens | Raises Dec 1, 2026 to $4.00 in / $20.00 out; $0.10 to $0.20 implicit cache |
| Gemini 4 Hydrogen | Free ($0) | $1.50 | $7.50 / 1M tokens | $0.075 / 1M implicit cache; unified 1M context window |
| Gemini 3.8 Flash | Free ($0) | $0.75 | $3.75 / 1M tokens | Raises Jan 1, 2027 to $1.50 in / $7.50 out |
| Nano Banana 2.1 Image | Not available | $1.50 | $30 / 1M ($0.0336 per 1K img) | 2K costs $0.0504, 4K costs $0.113; Batch cuts rates 50% |
| Gemini 3 Pro Image | Not available | $2.00 (~$0.0011/img) | $120 / 1M ($0.134 per 1K/2K) | 4K renders cost $0.24 per image |
| Gemini 3.8 Flash TTS | Free ($0) | $0.50 (text) | $9 / 1M audio ($0.00225 / 10s) | About $0.0135/min through Dec 31; doubles Jan 1, 2027 |
| Lyria 3.5 Music | Not available | Per request | $0.08 per full song | Legacy Lyria 3 30s preview clip runs $0.04 per song |

Image, voice, and Google Search grounding math
Three practical details in Google billing catch teams off guard during rollout when shifting workloads from ChatGPT subscriptions or standalone web generators.
First, Nano Banana 2.1 prices images by resolution inside the $30/1M image token pool: $0.0336 at 1K, $0.0504 at 2K, and $0.113 at 4K, while asynchronous Batch processing cuts those figures in half to $0.0168, $0.0252, and $0.0565. Upgrading from 2.1 to Gemini 3 Pro Image increases output spend roughly fourfold ($0.134 for 1K/2K, $0.24 for 4K).
Second, Gemini 3.8 Flash TTS runs on the Paid Tier at $0.50 per 1M text input tokens and $9.00 per 1M audio output tokens, which equals roughly $0.00225 per 10 seconds ($0.0135 per finished audio minute) through December 31, 2026 before stepping up to $0.0045 per 10 seconds ($0.027/minute) on January 1, 2027.
Third, Grounding with Google Search and Google Maps grants 5,000 free grounded requests per month (shared across Gemini 3 and newer models), then bills $14 per 1,000 search queries subject to a daily cap of 1,500 requests per day. A single prompt that triggers multiple sub-searches bills for each query executed.
Google AI Studio API vs flat subscriptions
Pay-as-you-go API billing keeps idle months at $0, which makes Google AI Studio far cheaper than flat monthly SaaS tools whenever generation happens in bursty weekly batches or inside scripted apps.
- For text-to-speech pipelines, Gemini 3.8 Flash TTS at $0.0135/minute competes directly with Murf Falcon 2.1 ($0.01/1K chars, ~$0.009/min) and Fish Audio S2.1 Pro ($0.015/1K bytes), undercutting ElevenLabs v4 API ($0.04 to $0.08/1K chars) by more than 70%.
- For commercial still images under NDA privacy rules, Nano Banana 2.1 ($0.0336 per 1K, $0.0168 Batch) costs under $3.40 for 100 private renders without paying $20/month for Ideogram Plus or $60/month for Midjourney Pro Stealth.
- For multi-model browsing and web UI character training without coding an integration, flat subscriptions like OpenArt Plus ($34/month commercial) or DeepAI Pro ($9.99/month) remain simpler to operate.
Run your expected monthly audio minutes and image counts through the Oqerune AI Cost Calculator to check where pay-per-unit API pricing beats flat monthly tiers.
Frequently asked questions
Is Google AI Studio free for commercial apps in 2026?
Google AI Studio lets developers prototype for free inside the Free Tier, but Free Tier inputs and outputs are used to improve Google products and lack higher production rate limits. Commercial apps needing privacy or flagship models like Gemini 4 Carbon and Nano Banana 2.1 must enable the Paid Tier.
How much does Nano Banana 2.1 cost per image in Google AI Studio?
Verified on 2026-11-14, standard Paid Tier Nano Banana 2.1 output ($30 per 1M image tokens) equals $0.0336 per 1K image, $0.0504 per 2K image, and $0.113 per 4K image. Batch API runs halve those prices to $0.0168 (1K), $0.0252 (2K), and $0.0565 (4K).
When do Gemini 4 Carbon and Gemini 3.8 Flash prices increase?
Introductory Paid Tier rates for Gemini 4 Carbon and Argon ($2.00 input / $10.00 output per 1M tokens) run through November 30, 2026 and adjust to $4.00 / $20.00 on December 1, 2026. Gemini 3.8 Flash ($0.75 / $3.75) and 3.8 Flash TTS ($9.00 audio output) double starting January 1, 2027.
How does Google Search grounding billing work in the Gemini API?
Gemini 3 and newer models share 5,000 free grounded Search and Maps requests per month on the Paid Tier, then bill $14 per 1,000 executed search queries under a daily limit of 1,500 requests per day.
What are thinking tokens and how do they affect my Google AI bill?
Gemini 4 and Flash models include internal reasoning steps and tool schemas inside total token billing, so complex chain-of-thought prompts increase output token charges even when the visible reply stays short.