Is the Gemini 4 API free? Only Hydrogen has a free tier
Hydrogen is the only Gemini 4 model on Google’s free API tier. See what free excludes, how to check your project and limits, and what to use instead.
Only one Gemini 4 model has a free API tier: Gemini 4 Hydrogen (gemini-4-hydrogen). Carbon, Neon and Argon are paid-tier only. Hydrogen’s free tier is rate-limited per Google Cloud project. Google does not publish fixed free-tier numbers for it. On that tier your prompts can be used to improve Google’s products, and Google Search grounding, Batch and Flex are not available.
We checked Google’s pricing page, rate-limits page, billing page, terms and troubleshooting guide on November 19, 2026. This article is based on that documentation. We did not run free-tier requests from a Google project, so we make no claims about real-world throughput. Ofox publishes this article, and the Ofox section near the end says what Ofox does and does not offer.
Which Gemini 4 models are free?
| Model | Model ID | Free tier on the Gemini API | Paid Standard price, USD per 1M input / output tokens |
|---|---|---|---|
| Gemini 4 Hydrogen | gemini-4-hydrogen | Yes, input and output listed as “Free of charge” | $1.50 / $7.50 |
| Gemini 4 Neon | gemini-4-neon | No | $3.00 / $15.00 |
| Gemini 4 Carbon | gemini-4-carbon | No | $2.00 / $10.00 through Nov 30, 2026; $4.00 / $20.00 from Dec 1 |
| Gemini 4 Argon | gemini-4-argon | No | $2.00 / $10.00 through Nov 30, 2026; $4.00 / $20.00 from Dec 1 |
Source: Gemini Developer API pricing, checked November 19, 2026. Google’s pricing page states this directly for three of the four models: Carbon, Argon and Neon each carry a note that they are “not available on the Free tier”. Hydrogen’s Standard table has a Free Tier column, and its input, output and context-caching rows all read “Free of charge”. The Hydrogen release note from November 12 also says it launched on the free tier.
The December price change for Carbon and Argon is covered in the Carbon and Argon December price change guide. It does not affect Hydrogen. If you are choosing between the models rather than just looking for a free one, the Gemini 4 model selection guide compares them by task.
Hydrogen’s free-tier rate limits
If you have seen fixed numbers such as “15 requests per minute, 1,500 per day” quoted for Gemini, they don’t come from Google’s current documentation for Hydrogen. As of November 19, 2026, the rate-limits page publishes free-tier numbers for only a few models. It does not publish any for Hydrogen.
The page lists exact tiers for Gemini Decision 1 (Free: 200 RPM, 50,000 TPM, 10,000 RPD) and Gemini Omni 1.2 Flash (Free: not supported). For every other model, including Hydrogen, it sends you to a per-project dashboard: aistudio.google.com/rate-limit. It also says that “Specified rate limits are not guaranteed and actual capacity may vary”. Hydrogen’s own developer guide describes the free tier only as “Free of charge”, marked as rate-limited, and links to the same page.
What Google does document about how the free limits work:
| Rule | What it means for a Hydrogen free project | Source |
|---|---|---|
| Limits are measured as RPM, input TPM and RPD | Going over any one of them returns a rate-limit error, even if you are under the others | Rate limits, “How rate limits work” |
| Limits apply per project, not per API key | Making a second key in the same project gives you no extra quota | Rate limits |
| RPD resets at midnight Pacific time | A daily quota error clears at 00:00 PT, not 24 hours after your first request | Rate limits |
| Spend-based limits: “N/A” for Free | The $10 / 10-minute spend cap applies from Tier 1 upward, not to free projects | Rate limits, “Spend-based rate limits” |
| Grounding with Google Search has its own 1,500 RPD default cap | Applies to paid use. Hydrogen’s free tier has no Search grounding at all (next section) | Rate limits; pricing |
| Failed requests (400/500) are not charged but still count against quota | Retry loops still use up your free allowance | Billing FAQ |
To see your actual numbers, open the rate-limit page in AI Studio while signed in to the account that owns the project and select the project behind your API key. The figures there are the only ones that apply to you. If you plan capacity, plan from that page and recheck it, because Google says limits change automatically as your tier and account status change.
When a free-tier request goes over a limit, the documented response is a 429 (rate_limit_exceeded, quota_exceeded or too_many_requests in the API error reference). A 429 whose quota details report a limit of 0 means something else. The Gemini free-tier limit-of-zero diagnosis explains how to read the quota metric in that case. For general retry behaviour, see the 429 retry guide.
What “free” does not include
Hydrogen’s free tier covers inference. Several features that sit next to it are paid-only, and the terms attached to free usage are different.
| Feature or condition | Hydrogen free tier | Hydrogen paid tier | Source |
|---|---|---|---|
| Input and output tokens (Standard) | Free of charge | $1.50 / $7.50 per 1M | Pricing, Hydrogen table |
| Context caching | Free of charge (implicit caching only) | $0.075 per 1M | Pricing |
| Grounding with Google Search | Not available | 5,000 free requests/month shared across Gemini 3+ models, then $14 per 1,000 | Pricing |
| Grounding with Google Maps | Not available | 5,000 free requests/month shared, then $14 per 1,000 | Pricing |
| Batch API (50% price) | Not available | $0.75 / $3.75 per 1M | Pricing, Batch tab |
| Flex | Not available | $0.75 / $3.75 per 1M | Pricing, Flex tab |
| Computer use tool | Not available (all models) | Billed as regular tokens | Pricing for tools |
| Code execution, URL context, File search | Free of charge (tools table) | Billed as tokens or embeddings | Pricing for tools |
| Your prompts and responses used to improve Google products | Yes | No (Standard, Batch, Flex, Priority; the invite-only Contributor tier says Yes) | Pricing; Terms |
A note on Search grounding. Google’s general “Pricing for tools” table gives the free tier for Google Search as “500 RPD free (limit shared for Flash and Flash-Lite)”. Hydrogen’s own pricing row says Search grounding is “Not available” on the free tier. We follow the model-specific row. If your agent depends on grounded web answers, plan on a paid project.
Data use: the biggest difference
Under the Gemini API terms (effective March 23, 2026), free usage counts as “Unpaid Services”. That covers both AI Studio and “unpaid quota in Gemini API”. For Unpaid Services, Google uses submitted content and generated responses to provide, improve and develop its products and machine-learning technologies. Human reviewers may read and annotate API input and output. Google says it disconnects the data from your account, API key and project before reviewers see it. The terms put the practical consequence plainly: “Do not submit sensitive, confidential, or personal information to the Unpaid Services.”
Paid usage reverses this. Google does not use prompts or responses to improve its products, processes them under its Data Processing Addendum, and logs them for a limited period for abuse detection and legal requirements. One detail in the billing FAQ catches people out. Your AI Studio prompts are treated under the paid terms as long as at least one API project on the account has billing enabled. The check is on the account, not the project you happen to be using.
Grounding with Google Search has its own retention rule. Google stores prompts, context and output for 30 days to create grounded results, on both tiers. That matters less for Hydrogen free usage, because Search grounding isn’t offered there.
Regions, age and the EEA/UK/Switzerland rule
- Supported regions. You can only use the API from a country or territory on Google’s available regions list (230 entries when we checked). Mainland China, Hong Kong and Russia are not on it. For Colab, the region of the Colab instance counts, not your own location.
- EEA, UK and Switzerland. The billing FAQ says the free tier is offered there. The terms add two conditions. If you make an API client available to users in those regions, you must use Paid Services only. And for users located there, the paid-tier data terms apply even to free usage.
- Age and purpose. You must be 18 or older. The terms describe the Gemini API and AI Studio as being for professional or business use, not consumer use.
Verification can block a free project
Some free projects must pass mobile-device verification before they can call the API. This applies to additional projects and to projects created outside AI Studio. An unverified project returns 403 with FREE_TIER_VERIFICATION_REQUIRED and a verification link (verification guide). Some projects are not eligible for the free tier at all and show Free tier unavailable. In that case a new key in the same project won’t help, because eligibility is checked per project. The only fix Google offers is linking billing, and doing that does not unlock free quota.
How to tell whether a project is on the free or paid tier
There are two places in Google AI Studio to look:
- Projects page (aistudio.google.com/projects). The Billing Tier column shows Free or a paid tier for each project. The Status column can show Set up billing (no billing account linked), Set up Prepay, No credits, Verification needed, Free tier unavailable or Payment needed.
- API keys page (aistudio.google.com/api-keys). Projects on billing show as “Paid” under Plan. Each key belongs to one project, so this tells you which tier a key will hit.
A paid status alone does not mean paid requests will succeed. New billing accounts default to Prepay. That means a minimum $5 credit purchase, and requests fail with 402 payment_required once the balance reaches $0 (billing). Also, models.list returning 200 OK does not prove anything. Google’s troubleshooting guide notes that this endpoint only checks that the key exists, not tier eligibility or billing. The real test is one short generateContent or Interactions call with the model you plan to use.
To go back to free, disable billing on the project. Some projects then need verification before the free tier works again.
What error do you get calling Carbon or Neon from a free project?
Google does not document a model-specific error for this case. The Carbon, Neon and Argon pricing entries say only that they are not available on the free tier. Google’s docs do describe project-level failures: 403 PERMISSION_DENIED with statuses such as Free tier unavailable or Verification needed, 402 payment_required for an empty Prepay balance, and 429 quota errors.
So match on the HTTP status and then check the project. Don’t match on an exact message string. The full symptom table, with fixes and verification steps, is in Gemini 4 API errors.
Free Gemini app access is not free API access
Google uses Gemini 4 models in consumer products too, and some of that usage is free. None of it gives you API quota.
- Gemini app. Google’s app release notes for November 17, 2026 say Tasks in Gemini is rolling out free of charge in the US, India, Canada, Australia, Japan and Korea, running on Hydrogen. An earlier entry (October 13, 2026) says Gemini 4 Argon in the app is for Google AI Pro and Ultra subscribers.
- Other surfaces. Hydrogen’s model card lists the Gemini app, Google Search, Google Sheets, Gemini for business and Google Antigravity as distribution channels alongside the API. Usage in those products follows the product’s own plan, not your API project.
- AI Studio with a consumer plan. Neon’s developer guide says Neon is available in Google AI Studio with AI Pro and Ultra plans. That is access inside AI Studio, not a free API quota for
gemini-4-neon. The API remains paid-tier only. - Gemini CLI. Google’s developer blog announcement set June 18, 2026 as the date Gemini CLI would stop serving requests for free Gemini Code Assist for individuals and for Google AI Pro and Ultra sign-in. It says the CLI remains accessible with “paid Gemini and Gemini Enterprise Agent Platform API keys”. See Gemini CLI free tier shutdown fixes.
Paying for Google AI Pro or Ultra does not turn a free API project into a paid one. API billing is a separate Cloud Billing account linked in AI Studio.
Do Google Cloud trial credits cover the Gemini API?
For new accounts, no. Google’s billing FAQ says the $300 Google Cloud Welcome credit and Free Trial credit cannot be used for the Gemini API or AI Studio. Since March 2026, Gemini API usage has been excluded from the Free Trial program. Accounts that received a Welcome credit before the change can spend the remainder on the Gemini API until it expires (90 days).
Some related points from the same page:
- A Google Cloud Free Trial account stays on the Gemini API free tier until you upgrade it.
- Prepay accounts must buy credits (minimum $5) before any eligible promotional Cloud credits can be applied. After that, the promotional credits are used first.
- Purchased Prepay credits expire after 12 months and are generally non-refundable.
When does paying for Hydrogen make sense?
Free is the right choice for evaluation, prototypes and personal tools with non-sensitive data. Move to paid when any of these apply: you need Search grounding, Batch or computer use; you handle data you can’t submit under the unpaid terms; you serve users in the EEA, UK or Switzerland; or your project’s free limits block the workload.
Our arithmetic, at Google’s paid Standard price (not a measured run). Take a request with 3,000 input tokens and 800 output tokens, thinking included. It costs 3,000 × $1.50/1M + 800 × $7.50/1M = $0.0045 + $0.0060 = $0.0105, or $10.50 per 1,000 requests. Two things move this a lot. Hydrogen defaults to high thinking, and thinking tokens are billed as output, so a lower thinking level often cuts cost more than shorter prompts do. Batch halves both rates if you can wait for results. Gemini 4 thinking levels covers how to pick the setting.
Alternatives if the free tier doesn’t fit
1. Paid Hydrogen on Google. This is the same model with the paid-tier data terms, higher limits and the paid-only features. Setup takes a billing account and a $5 Prepay purchase. The Hydrogen API setup guide walks through the Google route and a first request.
2. Other Google models that still have a free tier. As of November 19, 2026, the pricing page lists free Standard input and output for, among others, Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite, Gemini 2.5 Pro, Gemini 2.5 Flash and Gemma 4. The same data-use terms apply, and each model has its own limits on your AI Studio rate-limit page. Image models such as Nano Banana 2 and Nano Banana Pro have no free tier. If you are moving from 3.8 Flash, see the Gemini 3.8 Flash to Hydrogen migration guide.
3. Free tiers from other providers. The free LLM API tiers ranking compares Gemini against other providers’ offers for coding work.
4. Ofox (pay-as-you-go). Ofox is an API gateway, so it is not a free option. It charges per token from a prepaid balance, with no subscription. It has no free tier, no signup credit and no free trial. What it offers is one key and one bill across providers:
google/gemini-4-hydrogenandgoogle/gemini-4-argonare available on Ofox. Gemini 4 Carbon and Gemini 4 Neon are not (catalog checked November 19, 2026). For those two, use Google’s paid tier.- Hydrogen on Ofox supports the OpenAI-compatible (
/v1/chat/completions) and Gemini-native protocols. Current prices are on the Hydrogen model page and the Argon model page. - The default rate limit is 100 RPM per team, shared across all keys in the team, and can be raised by contacting Ofox. See the Ofox rate-limit docs for the current value.
- Ofox does not use prompts or responses sent through its standard API for model training. The upstream provider still receives the request content, under its own terms for the active route (Security & Data).
Google’s free quota cannot be spent through Ofox, and an Ofox balance cannot pay for Google’s own endpoint. They are separate accounts.
What we checked and what we didn’t
- Checked (November 19, 2026): Google’s pricing, rate-limits, billing, terms, available-regions, verification, API-error and troubleshooting pages, fetched live and compared with snapshots taken earlier that day. The Hydrogen pricing section matched the snapshot. We also checked Ofox’s catalog, model pages and docs.
- Not checked: actual free-tier RPM, TPM or RPD values. These are only visible per project inside AI Studio, and we did not use a Google free-tier project. We also did not test the exact error text a free project gets for Carbon or Neon.
- Arithmetic: the cost example is our calculation from list prices. It is not a measured bill.
Related Docs
Frequently Asked Questions
- How many free requests per day does Gemini 4 Hydrogen get?
- Google does not publish a fixed number. As of November 19, 2026, the rate-limits documentation lists free-tier figures for some models (for example Gemini Decision 1) but not for Hydrogen, and sends you to the per-project rate-limit page in AI Studio. Read the RPM, TPM and RPD shown there for your own project.
- Can I use the $300 Google Cloud free trial credit for the Gemini API?
- Not for accounts opened after March 2, 2026. Google’s billing docs say Welcome credits and Free Trial credits cannot pay for Gemini API or AI Studio usage. Older accounts that already hold the credit can spend what remains until it expires.


