Free Gemini API key: Google's free tier, and the prepaid route
A free Gemini API key exists, and Google issues it: a Google AI Studio key on a project with no billing linked runs on the Free tier, with per-model limits Google sets. This page states those rules as Google's rate-limit page gives them, then the prepaid alternative: 15 Gemini ids and the rest of the catalog on one RelaySwitch key, from $10.
Google's free tier, as Google states it
From the Gemini API rate-limits page, read 2026-10-31. Numbers come from that page only; where it defers to AI Studio, so does this one.
How it is granted
- The key comes from Google AI Studio. A project qualifies for the Free tier as an “Active project or free trial”; its billing tier cap is listed as N/A.
- Tier 1 needs you to “set up and link an active billing account”; the page adds that to leave the Free tier “you must first set up billing in AI Studio”, and that the upgrade typically takes effect instantly.
- Tier 2 follows $100 paid plus 3 days from the first payment; Tier 3, $1,000 plus 30 days.
How it is limited
- “Rate limits are applied per project, not per API key.” Two keys on one project share one quota.
- “Limits vary depending on the specific model being used.” The page shows your project's current per-model limits in AI Studio rather than in its text, so no per-model figure is typed here.
- “Requests per day (RPD) quotas reset at midnight Pacific time.” The spend-based limit per 10 minutes is N/A on Free, $10 on Tier 1.
No project quota, every Gemini id, failover
Three things a free-tier key cannot give you.
Limits you set
A key here has no project tier. Its requests-per-minute and tokens-per-minute limits are fields on the key, its budget is the hard stop, and a 429 means the limit you chose, not a quota that resets at midnight Pacific.
One key, the whole list
15 Gemini ids with a list price today, plus the Claude, GPT, DeepSeek, Grok and open models on the same key and the same balance. No project per developer, no second account.
Failover at the same price
A request that fails upstream is retried on another route at the same rate, and a failed request is not charged. The routing page has fallback lists and the auto router.
Every Gemini id, next to Google's list
USD per million tokens, read from the catalog at render. The last column is how many chat turns of 2,000 tokens in and 500 out a $10 top-up buys.
| Model | Input | Output | Google list in / out | Under list | Turns per $10 |
|---|---|---|---|---|---|
| Gemini Embedding 001gemini-embedding-001 | $0.15 | $0.00 | $0.15 / $0.00 | — | 33,333 |
| Gemini 2.5 Flash Litegemini-2.5-flash-lite | $0.075 | $0.30 | $0.10 / $0.40 | 25% | 33,333 |
| Gemini 3.1 Flash Litegemini-3.1-flash-lite | $0.15 | $0.90 | $0.25 / $1.50 | 40% | 13,333 |
| Gemini 3.1 Flash Lite Previewgemini-3.1-flash-lite-preview | $0.1875 | $1.125 | $0.25 / $1.50 | 25% | 10,666 |
| Gemini 3.5 Flash Litegemini-3.5-flash-lite | $0.225 | $1.875 | $0.30 / $2.50 | 25% | 7,207 |
| Gemini 3 Flash Previewgemini-3-flash-preview | $0.375 | $2.25 | $0.50 / $3.00 | 25% | 5,333 |
| Gemini 3.8 Flashgemini-3.8-flash | $0.4925 | $2.25 | $0.75 / $3.75 | 40% | 4,739 |
| Gemini 3.7 Flashgemini-3.7-flash | $0.4925 | $2.25 | $0.75 / $3.75 | 39% | 4,739 |
| Gemini 3.6 Flashgemini-3.6-flash | $0.4925 | $2.25 | $0.75 / $3.75 | 39% | 4,739 |
| Gemini 2.5 Flashgemini-2.5-flash | $0.30 | $2.50 | $0.30 / $2.50 | — | 5,405 |
| Gemini 4 Argongemini-4-argon | $0.60 | $3.00 | $2.00 / $10.00 | 70% | 3,703 |
| Gemini 3.5 Flash Highgemini-3.5-flash-high | $0.6353 | $3.8118 | $1.50 / $9.00 | 57% | 3,148 |
| Gemini 3.1 Pro Previewgemini-3.1-pro-preview | $1.20 | $7.20 | $2.00 / $12.00 | 40% | 1,666 |
| Gemini 2.5 Progemini-2.5-pro | $0.9375 | $7.50 | $1.25 / $10.00 | 25% | 1,777 |
| Gemini 3.5 Flashgemini-3.5-flash | $1.50 | $9.00 | $1.50 / $9.00 | — | 1,333 |
12 of 15 Gemini ids sit under Google's list price on output today. Google list: the catalog's record of Google's own price.
Call it
Python, with the OpenAI SDK
from openai import OpenAI client = OpenAI(base_url="https://api.relayswitch.io/v1", api_key="sk-relay-...")reply = client.chat.completions.create( model="gemini-3.5-flash", messages=[{"role": "user", "content": "Hello"}],)print(reply.choices[0].message.content)The Gemini ids answer the OpenAI chat completions format at https://api.relayswitch.io/v1/chat/completions, so the OpenAI SDK, Cursor, Cline, Open WebUI and the rest work as they are. Pick the id from the table; nothing else changes.
Hit a 429 on Google's tier already? The RESOURCE_EXHAUSTED fix page covers the retry rule and the structural fix.
Free Gemini API key: questions
Google's tier, and what prepaid changes.
Every Gemini id on one prepaid key
From $10 by card or in USDT, USDC, BTC, ETH, SOL. Limits you set, no project quota, no subscription.