oakmeet

Google AI Studio and Gemini 2.5 Pro: access, API key, first prompt, rate limits

Open AI Studio, create a key, send one checkable prompt with Gemini 2.5 Pro, then read the numbers on your project's Rate limits page.

You want to try Gemini 2.5 Pro in the before wiring it into your own script. The official entry is aistudio.google.com (Google AI Studio): sign in, create an API key, pick the model in Playground, send one checkable first prompt. Quotas and limits follow the numbers shown for your project in AI Studio.

The searchable job is Google AI Studio and Gemini 2.5 Pro: access, API key, first prompt, and rate limits. Capabilities and limits cite ai.google.dev only; trust the Rate limits page for your project, not third-party RPM tables.

2.5Pro endpoint gemini-2.5-pro
3RPM · TPM · RPD
1One checkable first prompt

Who should start in AI Studio

Good fit

Quick trials of Gemini 2.5 Pro, output comparisons, or an API key for a local prototype.

Not the finish line

Production keys stay on a server or in Secret Manager, not in a frontend repo.

StepWhereOfficial
Consoleaistudio.google.comapi-key docs
KeyDashboard → API Keysai.google.dev/.../api-key
LimitsRate limits pageai.google.dev/.../rate-limits
ModelPlayground / APIgemini-2.5-pro

One pass: sign in, key, first prompt

  1. 01
    Open Google AI Studio and sign in

    Visit aistudio.google.com and sign in with a Google Account. New users: official API key docs say accepting terms auto-creates a default Cloud project and key. Existing Cloud users import via Dashboard → Projects.

  2. 02
    Create or copy an API key

    Open API Keys from the Dashboard and click Create API key. New keys in 2026 default to auth keys; do not commit to Git or ship in client code. Set GEMINI_API_KEY or GOOGLE_API_KEY locally.

  3. 03
    Select Gemini 2.5 Pro in Playground

    Choose Gemini 2.5 Pro (endpoint gemini-2.5-pro). Official models page: complex tasks, deep reasoning, coding in the 2.5 family. Experimental/preview variants are usually stricter.

  4. 04
    Send one checkable first prompt

    Ask one verifiable thing, e.g. explain each JSON field in three sentences and list assumed types. For API calls, set model to gemini-2.5-pro.

export GEMINI_API_KEY="your_key"  # do not commit
curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-pro:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"Explain each key in three sentences"}]}]}'

How to read rate limits

What limits apply to
Official docs: per project, not per API key.
Three dimensions
RPM, TPM, RPD. Exceeding any returns 429 RESOURCE_EXHAUSTED.
When RPD resets
Midnight Pacific time.
Free vs paid tiers
Usage tiers table sets ceilings; check tier on AI Studio Projects.
Experimental / preview
Official docs: stricter limits; stable endpoint gemini-2.5-pro.

Sign in to AI Studio, create a key, send one checkable prompt with gemini-2.5-pro in Playground, then open your project's Rate limits page.

Not the same path as IDE Gemini

AI Studio in the browser is for model trials and API keys; Gemini Code Assist in VS Code follows Google Cloud org seats. To align on an API sketch or prompt draft, put bullets on one page. A short oakmeet room is enough — browser, space code, no account. Limits: eight people, about an hour. Compare Gemini Code Assist and Cursor Agent.

Do new users get an API key automatically?

Official API key docs: new users get a default Cloud project and API key after accepting terms in AI Studio. Existing Cloud users import a project first.

Does 429 always mean RPM?

No. Official rate-limits and troubleshooting list RPM, TPM, RPD, and (on paid tiers) rolling 10-minute spend limits as separate 429 triggers.

Can I put the API key in a web frontend?

Official security guidance says not to compile keys into production client code; use a backend proxy or Secret Manager. Leaked keys may be blocked.

Gemini 2.5 Pro vs 2.5 Flash?

Official models page: Pro for complex tasks and deep reasoning; Flash for low-latency, high-volume price-performance. Compare Flash limits for high-frequency calls.

Start a room