Gemini's free tier got smaller this year. Here's what's left.
Google tightened the Gemini API free tier hard in 2026. Free access to Pro-tier models ended on April 1, and rate limits keep shifting without much warning. If you're building something on Gemini and don't want to pay yet, here's what the free tier still covers and how to stretch it.
What you get on the free tier now
As of this year, only Flash and Flash-Lite models remain on the free tier. Gemini Pro requires a billing account, full stop. Google also no longer publishes fixed request-per-day numbers for every model since it adjusts them without notice (it cut free quotas by 50 to 80 percent back in December 2025). Your actual current limits show up live in AI Studio, not in a static docs page.
Rough numbers reported by developers right now: Flash and Flash-Lite sit around 5 to 15 requests per minute, up to roughly 1,000 requests a day, though this varies by account and region. Flash-Lite gets a higher requests-per-minute ceiling than Flash, which matters if your workload comes in bursts.
Google also refreshed the Flash line on July 21 with [Gemini 3.6 Flash and 3.5 Flash-Lite](https://questloops.com/blog/gemini-3-6-flash-explained-pricing-benchmarks-and-what-actually-changed), so if you're on the free tier through the API, the model behind your requests already changed even if the free-tier limits above haven't.
One more thing worth knowing before you build anything real on it: free-tier prompts can be used to improve Google's models. The paid tier is not used for training. If that matters for anything you're sending through the API, it's worth reading before you commit.
Free tier vs paid tier
| Free tier | Paid tier | |
|---|---|---|
| Models available | Flash, Flash-Lite only | Flash, Flash-Lite, Pro |
| Rate limits | ~5 to 15 RPM, ~1,000 requests/day (varies) | Higher, scales with billing |
| Price | $0 | Starts at $0.075 per million input tokens |
| Data use | May be used to improve Google's models | Not used for training |
| Credit card required | No | Yes |
AI Studio itself, the web interface for testing prompts, stays free with no credit card needed in every supported country, even though the API free tier around it has shrunk.
How to make the most of the free tier
- Grab an API key through Google AI Studio, no billing setup required for the free path.
- Default to Flash-Lite for anything bursty. Its higher RPM ceiling means fewer 429 errors when requests cluster.
- Check your live quota in AI Studio rather than trusting any number you read online, including this article. Google adjusts these without notice.
- If you need Pro-level reasoning, budget for it. There's no free path to Pro anymore, and workarounds that claim otherwise tend to be rate-limit tricks that stop working within weeks.
How this compares to other free AI options
If Gemini's shrinking free tier is pushing you to look elsewhere, [Claude's free tier](https://questloops.com/blog/use-claude-ai-free) and the free version of [ChatGPT](https://questloops.com/blog/how-to-get-chatgpt-plus-free) are both still worth checking, and our [roundup of what's still free across AI tools in 2026](https://questloops.com/blog/ai-student-discounts-in-2026-what-s-still-free-vs-what-already-expired) covers the broader landscape if you're a student.
The takeaway
Gemini's free tier is real but narrower than it was a year ago: Flash and Flash-Lite only, live-adjusted rate limits, and your prompts feeding Google's training data unless you pay. It's genuinely fine for prototyping and small projects. Just don't build anything you depend on around Pro access without a billing account behind it.
