import { GoogleGenAI } from \"@google\/genai\";<\/p>\nconst ai = new GoogleGenAI({ apiKey: \"YOUR_API_KEY\" });<\/p>\n
async function run() {
console.log(response.text);
run();
REST:<\/strong><\/p>\ncurl \"https:\/\/generativelanguage.googleapis.com\/v1beta\/models\/gemini-3-pro-preview:generateContent\" \\If you get a response, you’re set up correctly.<\/p><\/blockquote>\n
Free Tier vs Paid Access<\/h3>\n\n<\/div>\n<\/figure>\n
What’s included in the free tier:<\/strong><\/p>\nGoogle AI Studio gives you free access to Gemini 3 Pro with rate limits. You can prototype, test, and build small applications without paying anything.<\/p>\n
Rate limits (approximate):<\/strong><\/p>\n\n5-10 requests per minute (RPM)<\/li>\n 250,000 tokens per minute (TPM)<\/li>\n 50-100 requests per day (RPD)<\/li>\n<\/ul>\nThese limits are enough for development and testing, but not for production traffic.<\/p>\n
When to upgrade to paid access:<\/strong><\/p>\nYou need paid access when:<\/p>\n
\nYour traffic exceeds free tier limits<\/li>\n You need higher throughput<\/li>\n You’re deploying to production<\/li>\n You require enterprise features (Vertex AI)<\/li>\n<\/ul>\nCost optimization tip:<\/strong><\/p>\nIf you’re staying on the free tier, create multiple Google accounts (each gets its own key) and rotate between them. Common practice among developers staying within free limits.<\/p>\n
Understanding Thinking Levels (Your Speed vs Intelligence Control)<\/h2>\n Gemini 3 introduces a new way to control how much reasoning the model uses.<\/p>\n
What Thinking Levels Actually Do<\/h3>\n The thinking_level<\/code> parameter controls the maximum depth of the model’s internal reasoning process before responding.<\/p>\nThink of it like this: Gemini 3 can think longer about hard problems and respond quickly to easy ones. The thinking level sets the upper limit on how much time it takes to reason.<\/p>\n
Important:<\/strong> Gemini 3 uses dynamic thinking by default. Even on high<\/code> thinking level, it won’t waste time on simple questions. It only uses deep reasoning when the prompt needs it.<\/p>\nChoosing The Right Thinking Level<\/h3>\n low<\/code> – Minimize latency and cost<\/strong><\/p>\nUse when:<\/p>\n
\nSimple instruction following<\/li>\n Chat and conversational AI<\/li>\n High-throughput applications<\/li>\n Speed matters more than depth<\/li>\n<\/ul>\nExample: Customer support chatbot answering FAQs<\/p>\n
high<\/code> – Maximize reasoning depth (default)<\/strong><\/p>\nUse when:<\/p>\n
\nComplex problem-solving<\/li>\n Multi-step reasoning<\/li>\n Code generation with intricate logic<\/li>\n Accuracy matters more than speed<\/li>\n<\/ul>\nExample: Debugging race conditions in multi-threaded code<\/p>\n
medium<\/code> – Coming soon<\/strong><\/p>\nWill offer a balance between the two. Not available at launch.<\/p>\n
Default behavior:<\/strong><\/p>\nIf you don’t specify thinking_level<\/code>, Gemini 3 Pro defaults to high<\/code>. For most applications, this is what you want.<\/p>\nCode Examples For Each Level<\/h3>\n Python (setting to low):<\/strong><\/p>\nfrom google import genai<\/p>\nclient = genai.Client()<\/p>\n
response = client.models.generate_content(
print(response.text)
Python (default high):<\/strong><\/p>\nresponse = client.models.generate_content(JavaScript:<\/strong><\/p>\nconst response = await ai.models.generateContent({REST:<\/strong><\/p>\ncurl \"https:\/\/generativelanguage.googleapis.com\/v1beta\/models\/gemini-3-pro-preview:generateContent\" \\Common Mistakes to Avoid<\/h3>\n Don’t use both thinking_level<\/code> and thinking_budget<\/code><\/strong><\/p>\nThe old thinking_budget<\/code> parameter is still supported for backward compatibility, but using both in the same request returns a 400 error.<\/p>\nIf you have old code using thinking_budget<\/code>, migrate to thinking_level<\/code> for better performance.<\/p>\nDon’t set low<\/code> for everything<\/strong><\/p>\nIt’s tempting to use low<\/code> thinking for speed, but complex tasks need deep reasoning. If your results are worse after switching to low<\/code>, try high<\/code> instead.<\/p>\nMedia Resolution: Getting The Best From Images and Video<\/h2>\n Gemini 3 gives you granular control over how it processes visual content through the media_resolution<\/code> parameter.<\/p>\nWhy Media Resolution Matters<\/h3>\n\n<\/div>\n<\/figure>\n
Higher resolution means:<\/p>\n
![](\"https:\/\/godofprompt.ai\/blog\/wp-content\/uploads\/2026\/05\/6956a4bc5477bb4ad1950b34_69130b993ec445432c7281c9_690de01d9ed6b3cf56ad762e_Design_1_1200x628.webp\")
<\/div>\n![\"__wf_reserved_inherit\"](\"https:\/\/godofprompt.ai\/blog\/wp-content\/uploads\/2026\/05\/6956a53e19d89bf239ed5eda_691d8c43d395e3017fa5b920_Context-Window-and-Pricing.avif\")
<\/div>![\"__wf_reserved_inherit\"](\"https:\/\/godofprompt.ai\/blog\/wp-content\/uploads\/2026\/05\/6956a53e19d89bf239ed5ed4_691d8c7719701165bde3edbf_Step-2-Generate-your-API-key.avif\")
<\/div>![\"__wf_reserved_inherit\"](\"https:\/\/godofprompt.ai\/blog\/wp-content\/uploads\/2026\/05\/6956a53e19d89bf239ed5edd_691d8ccb64fcd99e5945b478_Free-Tier-vs-Paid-Access.avif\")
<\/div>\n<\/figure>\n![\"__wf_reserved_inherit\"](\"https:\/\/godofprompt.ai\/blog\/wp-content\/uploads\/2026\/05\/6956a53e19d89bf239ed5ed7_691d8cf952b45ad356ef0dfb_Why-Media-Resolution-Matters.avif\")
<\/div>\n<\/figure>\n