Gemini 3.1 Flash Image Free Tier: API Is Paid, AI Studio Can Test It

As of June 13, 2026, gemini-3.1-flash-image does not have an official Gemini Developer API Free Tier row.

Google AI Studio can still be useful for trying Nano Banana 2 in the browser, but that does not create free production API quota for your backend. For app or service traffic, plan around the paid Developer API rows unless Google's pricing page changes.

The current model ID is gemini-3.1-flash-image. Older gemini-3.1-flash-image-preview examples should be treated as legacy context rather than the current code contract.

Route answer: use AI Studio when you only need browser testing, use the paid Developer API for synchronous backend traffic, use Batch only when asynchronous paid jobs fit, and keep Gemini Apps limits separate from API quota.

Stop rule: do not ship production traffic because a snippet says the model is "free." Check the Google pricing row for model eligibility and AI Studio for the active project, model, and tier limits behind your key.

Quick Answer

The short version is route-specific. "Free" can mean a browser test in AI Studio, a consumer app feature, a different Gemini model row, or a provider-owned promotion. It does not automatically mean free production API calls for Gemini 3.1 Flash Image.

For broader Gemini model free-tier questions, use the Gemini API free tier limits guide. For this model-specific image workflow, the paid-only pricing row is the controlling fact.

Pick The Right Access Route

The safest way to plan Gemini 3.1 Flash Image access is to choose the route before you think about code or quota.

That split prevents the most expensive mistake: seeing AI Studio or Gemini Apps availability and assuming a server can call the same image model for free. The browser route proves you can try the model. The API pricing row decides whether your backend traffic has a free tier.

Official API Pricing Is Paid-Only For This Model

Google's pricing page is the source of truth for the Developer API free-tier question. As checked on June 13, 2026, the gemini-3.1-flash-image Standard row lists Free Tier input and output as not available. Its paid output image price is $60 per 1 million image tokens.

Google's own examples turn that paid Standard price into these approximate per-image costs:

Batch is also paid-only for this model, but the output image price is lower at $30 per 1 million image tokens. Google's example costs are:

Those numbers should be treated as Google Developer API pricing examples, not universal market prices. They also need a date because model rows, preview labels, and billing surfaces can change. Before a launch, open the pricing page again and confirm the exact row for gemini-3.1-flash-image.

Batch deserves a careful reading. It can reduce cost when the job is asynchronous, but it does not create a free lane. Use it for workloads such as catalog refreshes, background creative variants, or scheduled asset generation where delayed completion is acceptable. Use Standard when the user is waiting for the image in an app session.

Use The Current Model ID

For new API work, use:

hljs txtCopy
gemini-3.1-flash-image

Do not start new code with:

hljs txtCopy
gemini-3.1-flash-image-preview

Google's changelog says the GA gemini-3.1-flash-image and gemini-3-pro-image versions were released on May 28, 2026. The same changelog marks the preview versions as deprecated with shutdown on June 25, 2026. That makes the old preview string useful only when you are reading older examples, auditing existing code, or preserving URL continuity for a refreshed page.

The image-generation docs also map Nano Banana 2 to gemini-3.1-flash-image. Use the display name when talking to humans and the exact model ID in code, logs, allowlists, and billing dashboards.

hljs pythonCopy
from google import genai

client = genai.Client()

response = client.models.generate_content(
    model="gemini-3.1-flash-image",
    contents="Create a clean product mockup on a neutral background.",
)

The operational check is simple: if a repository, tutorial, or dashboard still uses the preview suffix, verify whether it is historical context or active code. Active production code should migrate before the shutdown date.

How To Check Live Quota Before You Build

Google's rate-limit docs describe limits in RPM, TPM, and RPD: requests per minute, tokens per minute, and requests per day. They also frame limits by project, model, and tier. That means a static number from an older article is not enough for Gemini 3.1 Flash Image planning.

Use this order before a prototype becomes a customer-facing feature:

1. Open AI Studio with the account that owns the API key or project.
2. Select the same project your code will use.
3. Confirm the model ID is gemini-3.1-flash-image.
4. Check whether the project is on a free, billed, or higher usage tier.
5. Record the active RPM, TPM, RPD, and any UI caveats shown for that project.
6. Re-check before a demo, launch, pricing change, migration, or traffic spike.

The API key itself is not the quota owner. The project behind the key is the practical owner of limits, billing state, and usage reporting. Creating extra keys in the same project is useful for rotation and environment separation, but it is not a reliable quota multiplier.

If you hit 429 or RESOURCE_EXHAUSTED, start by checking the project, model, and tier in AI Studio. Then reduce concurrency, add backoff, shorten prompts where possible, cache repeated outputs, or move the workload to the paid lane that matches normal traffic. For a deeper troubleshooting path, use the Gemini image generation rate limit guide.

When AI Studio Is Enough

AI Studio is the right first move when the goal is to learn whether Gemini 3.1 Flash Image can produce the kind of image you need. It keeps the early test close to Google's browser surface and avoids prematurely designing backend billing, storage, retries, and moderation around an unproven visual workflow.

Use AI Studio for:

• comparing prompt wording and reference-image behavior
• testing whether Nano Banana 2 fits the style or editing task
• capturing examples for an internal design review
• deciding whether a paid backend implementation is worth building

Do not use AI Studio as proof of production API entitlement. A browser generation can be available while the official Developer API row remains paid-only. A consumer or browser limit can also change without becoming a backend SLA.

The handoff point is workload responsibility. Once users, jobs, retries, logs, billing, data retention, or uptime expectations matter, plan around the Developer API route. If the job can wait, compare Standard with Batch. If the job is interactive, price Standard and design for rate-limit recovery.

Gemini Apps Are A Separate Consumer Surface

Gemini Apps can expose image generation to consumers under app-plan rules and compute-based limits. That is useful for personal work, but it does not answer the Developer API free-tier question.

Keep these boundaries straight:

This separation also prevents bad support decisions. If a user says image generation worked in Gemini Apps but failed in an API integration, do not debug it as one shared limit. Check the API model ID, pricing eligibility, project tier, billing state, and rate-limit dimension.

Decision Rules For Developers

Use this working rule set:

Before choosing a third-party route, verify current model coverage, cost basis, billing unit, failure policy, output ownership, data handling, latency, rate limits, and support path directly from that provider. Without that evidence, the safer published advice is to stay with the official route decision: AI Studio for testing, paid Developer API for production traffic, and Batch only when async delivery fits.

For model choice rather than free-tier status, use the dedicated Gemini 3 Pro Image vs Gemini 3.1 Flash Image comparison. The free-tier decision should not be mixed with the Pro-vs-Flash quality decision.

Migration Checklist From Preview-Era Examples

Use this checklist when refreshing code or docs that still mention gemini-3.1-flash-image-preview:

The goal is not just syntactic cleanup. A stale model string can produce deprecation failures, stale pricing language can mislead budgeting, and mixed route wording can make support teams debug the wrong surface.

FAQ

Does Gemini 3.1 Flash Image have a free API tier?

No for the official Gemini Developer API row checked on June 13, 2026. Google lists Free Tier as not available for gemini-3.1-flash-image Standard and Batch image rows.

Can I use Google AI Studio for free image testing?

AI Studio can be used to try the model in the browser, subject to the active project and account limits shown there. Treat that as a testing route, not as free production API quota.

Is Nano Banana 2 the same as Gemini 3.1 Flash Image?

Yes. Google's image-generation docs map Nano Banana 2 to the current gemini-3.1-flash-image model ID.

Should I still use gemini-3.1-flash-image-preview?

Not for new production work. Use gemini-3.1-flash-image. The preview string belongs in legacy code review, older examples, migration notes, or historical URL context.

Is Batch free because it is cheaper?

No. Batch is a lower-cost paid asynchronous lane. It can reduce image output cost when latency does not matter, but it is not a Free Tier row.

Are Gemini Apps limits the same as API limits?

No. Gemini Apps are consumer app surfaces with their own compute-based limits and feature rules. Developer API quota belongs to the project, model, and tier behind the API key.

Where should I check my exact limits?

Use AI Studio for the active project behind your key. Check the model ID, project, tier, RPM, TPM, RPD, and any visible account or billing caveats before shipping.

What if I need a broader Gemini free-tier map?

Use the Gemini API free tier limits guide for model-family and project-quota questions. Use this model-specific route when the question is Gemini 3.1 Flash Image / Nano Banana 2 access.