AI API Free Tiers Compared

Each one takes something in return. Google may use content sent on the Gemini free tier to improve its products, while paid tier content stays out of that. Gemini 3.1 Pro Preview has no free tier at all. OpenRouter counts requests across every free model together, and a negative account balance blocks even free requests until you top up. On Groq, the Developer plan rather than the Free plan is what turns on Batch and Flex processing, so the free tier is for trying, not shipping.

None of this replaces credits for production traffic. Gemini quotas apply per project rather than per API key and daily counts reset at midnight Pacific time, and the set of OpenRouter free models rotates without notice. xAI doesn’t publish a data-sharing credit offer anywhere official, and OpenAI and Anthropic publish no free API tier at all, so none of the three appears in the picks below. Five dollars on either of those is the real entry price.

How to get the most from API free tiers

Read the data use column first

Google marks free tier Gemini models as content used to improve our products and paid tier models as not. If your prompts contain customer data, a contract or anything under NDA, the free tier is the wrong place for them regardless of cost. Linking a billing account changes the data terms as well as the limits, so do it before the first real request, not after. The column is easy to miss and expensive to ignore.

Spend $10 once on OpenRouter

OpenRouter raises the free model daily limit from 50 to 1,000 requests after you’ve purchased at least $10 in credits, and the threshold counts credits purchased over the life of the account rather than the current balance. A single $10 top-up therefore buys a twenty-fold daily allowance on free models for as long as the account exists. That’s the best deal on this page, and we’d do it the day you sign up.

Keep the OpenRouter balance positive

A negative balance on OpenRouter returns errors on every request, including free models, until you add enough credit to get back above zero. Failed requests on free endpoints still count against the daily allowance, so a retry loop against a busy model can burn the whole 50 before lunch. Add backoff, cap retries at two, and alert on the balance. Treat the free tier like it costs money, because the quota does.

Count Gemini quotas per project

Gemini rate limits are enforced per Google Cloud project, not per API key, so two keys in one project share one quota. Requests per day reset at midnight Pacific time. Built-in tools carry their own caps; Grounding with Google Search on Gemini 3 and newer models defaults to 1,500 requests per day per project. Separate a demo project from the one your app uses, or a conference demo can empty your app’s daily quota.

Skip the newest Pro model for free

Gemini 3.1 Pro Preview isn’t available on the free tier, while Gemini 2.5 Pro, 2.5 Flash, 2.5 Flash-Lite and 3 Flash Preview are free of charge. For reasoning-heavy prototypes, 2.5 Pro remains the free option; for everything else, 3 Flash Preview is the newer free choice. Test on the free models and reserve 3.1 Pro for the paid tier once the product works. You won’t miss it during prototyping.

Budget the first paid tier

When you outgrow the free tier, Gemini paid prices are $0.30 per million input tokens and $2.50 per million output tokens on 2.5 Flash, and $0.10 and $0.40 on 2.5 Flash-Lite. Tier 2 opens after $100 in payments plus 3 days and Tier 3 after $1,000 plus 30 days. A few dollars of real spend on a linked account raises limits far more than any workaround on the free tier, and it’s cheaper than the engineering time.