Free AI API Keys and Tokens: Official Free Tiers (Nov 2026)

Where can I get a free AI API key?

You can get a free AI API key directly from Google AI Studio for the Gemini API, from the Groq console, from a Cloudflare account for Workers AI, from OpenRouter and from NVIDIA's developer program, each with its own limits as of 13 November 2026 ^{[1][5][6][7][8]}. These are the providers whose own pricing or documentation pages describe a free tier that works through the API, not only in a chat app.

Google describes its free tier as being for developers and small projects getting started ^[1], and NVIDIA's free endpoints are for prototyping ^[8]. Each provider sets rate limits, restricts which models are free, or both, and Google says content sent on its free tier is used to improve its products ^[1]. Check each provider's terms before building a product with real users on a free tier.

Some names that used to appear on free-API lists no longer belong there. GitHub's documentation says GitHub Models was fully retired on 30 July 2026, including its inference API ^[9]. Mistral's documentation says its Free mode "doesn't include API access" and that API keys need billing activation and pay-as-you-go ^[10]. Hugging Face's pricing page lists free users as receiving no monthly credits for its Inference Providers ^[11].

For a broader comparison of the big chatbots and their paid prices, see the Pouyam guide ChatGPT vs Gemini vs Claude in 2026: Models, Prices and Key Differences.

Is the Gemini API free?

Yes, the Gemini API has an official free tier with free input and output tokens on a limited set of models, as of 13 November 2026 ^[1]. Google's pricing page lists free-tier access for models including Gemini 4 Hydrogen, Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 2.5 Pro and Gemini 2.5 Flash, plus Gemma 4 and embedding models ^[1]. The same page marks Gemini 2.5 Pro and 2.5 Flash as deprecated: from 1 December 2026 requests to them are routed automatically to Gemini 3.8 Flash and Gemini 3.5 Flash-Lite respectively ^[1].

Some models are paid only. As of 13 November 2026 the pricing page shows no free tier for Gemini 4 Carbon, Gemini 4 Argon, Gemini 3.1 Pro Preview or the Nano Banana image models ^[1]. Grounding with Google Search is listed as free only for Gemini 2.5 Flash and 2.5 Flash-Lite, up to 500 requests per day shared between them; for Gemini 3 and newer models and for Gemini 2.5 Pro the free tier shows it as not available ^[1].

The trade-off is data. The free tier is marked "Content used to improve our products", while the paid tier is not ^[1]. Do not send private, customer or confidential data through a free Gemini key.

Limits apply per project, not per API key, and the daily request count resets at midnight Pacific time ^[2]. Google does not publish fixed per-model numbers on the page; it says your active limits are shown in AI Studio ^[2]. Availability is country-based: Google's list of supported regions includes Türkiye but not Iran, as of 13 November 2026 ^[3].

Which AI APIs have a free tier in 2026?

Five providers' own pages describe a free API allowance as of 13 November 2026: Google Gemini, Groq, Cloudflare Workers AI, OpenRouter and NVIDIA ^{[1][5][6][7][8]}. They work differently, so compare them by what you need.

• Groq hosts open models. Its free plan allows 30 requests per minute, 1,000 per day, 8,000 tokens per minute and 200,000 tokens per day on openai/gpt-oss-120b, openai/gpt-oss-20b and qwen/qwen3.8-27b, and limits apply per organization (as of 13 Nov 2026) ^[5]. Speech-to-text with whisper-large-v3 allows 20 requests per minute and 2,000 per day ^[5].
• Cloudflare Workers AI gives 10,000 Neurons per day free on both its Free and Paid plans; Neurons are Cloudflare's unit of compute. The allowance resets at 00:00 UTC, and on the Paid plan extra use costs $0.011 per 1,000 Neurons (as of 13 Nov 2026) ^[6].
• OpenRouter offers free model variants. Models with a :free suffix are limited to 20 requests per minute and 50 per day if you have bought fewer than 10 credits, or 1,000 per day once you have bought at least 10 (as of 13 Nov 2026) ^[7].
• NVIDIA says members of its Developer Program have free access to its NIM API endpoints for prototyping; production use needs an NVIDIA AI Enterprise licence (as of 13 Nov 2026) ^[8].

The table below lines up the limits with the date each was checked.

How do I get a free API key safely?

Create the key yourself, in the provider's own console, while signed in to your own account; never take a key from someone else ^[4][7].

1. Go to the provider's official site by typing the address or using a bookmark, not through an ad or a link in a chat group.
2. Sign up with your own email and turn on two-factor authentication for the account.
3. Create a key in the console. For Gemini, Google says that since 28 May 2026 new keys made in AI Studio are created as auth keys, which are restricted to the Gemini API by default ^[4].
4. Store the key in an environment variable or a secrets manager, not in your code, a public repository or a screenshot.
5. Set a spending limit or keep billing off if you only want the free tier. OpenRouter warns that a negative credit balance can cause errors even on free models ^[7].
6. Delete keys you no longer use, and create a new one at once if a key may have leaked.

Google says auth keys come with fast leaked-key enforcement that stops the use of keys its systems detect as exposed ^[4]. That protects you, but it also means a key you paste into a public place may stop working.

Are free AI token websites a scam?

Treat any website, Telegram channel, app or browser extension that offers free AI tokens, shared API keys or "unlimited" access outside the official providers as a security risk and do not use it. The official documentation shows how the legitimate route works: keys are created for your own account in the provider's console ^[4], and limits are tied to a project or organization ^[2][5], or are governed globally by the provider ^[7].

OpenRouter's documentation is explicit: creating extra accounts or keys does not raise your free limits, because limits are global ^[7]. A key that has leaked can be stopped: Google says auth keys come with enforcement that quickly stops the use of keys its systems detect as leaked ^[4]. A shared or resold key therefore may stop working at any time.

Risks to think about before using any such offer:

• Your data: requests sent through someone else's key or proxy pass through their servers, so they can read your prompts and files.
• Your money and device: none of the providers above asks for a card payment, a crypto payment or a software download to hand out a free key; they create keys in your own account console ^[4]. A site that asks for these is not following that route.
• Your account: only sign in on the provider's own address, typed in or bookmarked.

The Pouyam guide Is ChatGPT Safe for Your Privacy? What AI Chatbots Do With the Data You Share explains what happens to data you send to AI services.

What happens when I hit the free limit?

When you exceed a free limit, the API refuses further requests, usually with an HTTP 429 "too many requests" error, until the limit resets ^[2][5]. Gemini's daily count resets at midnight Pacific time ^[2], and Cloudflare's daily Neurons reset at 00:00 UTC ^[6].

Practical ways to stay within limits: cache answers you reuse, send shorter prompts, use a smaller model for simple tasks, and add a short wait and retry when you receive a 429. On OpenRouter, a negative credit balance can cause errors even on free models, so check your balance first ^[7].

If your project grows beyond a free tier, move to the provider's paid plan rather than looking for workarounds. On Gemini, content sent on the paid tier is not used to improve Google's products ^[1].

Will free AI API tiers last?

Free tiers change often, so check the provider's own page before you rely on one; this guide's figures are dated 13 November 2026. In 2026 alone, GitHub retired its Models service on 30 July ^[9], and Google changed how new Gemini keys work from 28 May ^[4]. Cloudflare's pricing page was last updated on 12 November 2026 ^[6].

Examples in this guide: GitHub closed its service ^[9], Mistral's Free mode excludes the API ^[10], Google lists some newer models as paid only and is retiring Gemini 2.5 Pro and 2.5 Flash on 1 December 2026 ^[1], and Google's free tier uses content to improve its products ^[1]. If a free tier disappears, your code should fail safely, so keep the provider and model name in a config setting you can change in one place.