Google AI Studio
Browser workspace for testing Gemini prompts and parameters, then exporting an API key and code.
Quick facts
| Pricing | Freemium |
|---|---|
| Free tier | Yes — free usage with rate limits on the hosted models |
| Platforms | Web, API |
| Developer | |
| Website | https://aistudio.google.com/ |
What Google AI Studio is
Google AI Studio is a browser environment for working with Gemini models before writing production code. You write a prompt, adjust parameters, attach images or files and compare outputs, then take the generated code snippet into an application with an API key. It also supports building simple apps directly in the browser and saving prompts for reuse. The appeal is the short distance from experiment to API call: the same model and parameters carry over, so a working prompt is a few lines from being deployed. The limits are scope and quotas: it is a developer tool rather than an end-user product, free usage is rate-limited, and features arrive at different times from the consumer Gemini app. Prompt history can be compared, so a change in wording can be traced to a change in output. A structured output mode helps when the response has to fit a schema rather than prose.
Best for
- Testing a prompt before writing any code
- Getting an API key and a working snippet quickly
- Developers evaluating model behaviour on their own data
Not ideal for
- End users who want a finished assistant
- Production workloads without quota planning
Pros
- Prompt, parameters and code export in one screen
- File and image input for testing multimodal prompts
- Free tier for evaluation
Cons
- Rate limits apply to free usage
- It is a developer surface, not a consumer product
- Feature timing differs from the consumer app
Pricing in detail
Prompt testing and basic use are free with rate limits per model. Production use is billed through the Gemini API per token, with different rates for input, output and model tier, and volume discounts on higher tiers. Because AI Studio itself has no subscription, the cost question is entirely about API usage, which depends on how much text and how many images each request carries.