Overview
- Tool name: Gemini 3 (gemini-3.1-pro-preview / gemini-3-flash-preview and more)
- Developer: Google
- Official website: https://ai.google.dev
- Category: AI Models
- Status: Preview
Gemini 3 is Google's third-generation frontier family, whose current members (all in preview) include:
- Gemini 3.1 Pro (gemini-3.1-pro-preview): the flagship tier, 1M-token context (64K output), built for multimodal understanding, agentic workflows and coding (vibe coding); 54.2% on Terminal-Bench 2.0
- Gemini 3 Flash (gemini-3-flash-preview): the value tier - frontier-class intelligence at less than a quarter of Pro's price, with GPQA Diamond 90.4% and SWE-bench Verified 78%; it is the default model in Google Search's AI Mode
- Gemini 3.1 Flash-Lite (gemini-3.1-flash-lite): the high-throughput tier at $0.25/$1.50
- Gemini 3 Pro Image / Nano Banana Pro and Nano Banana 2: the image-generation tiers, with text priced as 3.1 Pro
The family ships with new capabilities: the thinking_level parameter (minimal/low/medium/high to trade reasoning depth against latency), context caching (up to 90% savings), the Batch API at half price, and Google Search / Maps grounding (5,000 free requests per month, shared).
API pricing (official, per million tokens)
- Gemini 3.1 Pro: $2 input / $12 output (prompts up to 200K tokens); $4/$18 beyond 200K; context caching $0.20
- Gemini 3 Flash: $0.50 input ($1.00 audio) / $3.00 output; caching $0.05
- Gemini 3.1 Flash-Lite: $0.25 input ($0.50 audio) / $1.50 output
- Batch at half price; a rate-limited free tier is available in AI Studio
Pricing verified against Google's official pricing page ai.google.dev/pricing and the Gemini 3 developer guide (October 2026).
Ecosystem
Gemini 3 is embedded across Google's stack: AI Studio, Gemini CLI, Android Studio, Google Antigravity (the agentic development platform), and Vertex AI for enterprises; third-party tools like Cursor, Cline and JetBrains support it directly.
Who it is for
- Budget-sensitive frontier needs: 3.1 Pro is the cheapest frontier flagship of 2026
- High-throughput production: Flash/Lite plus caching and Batch compress costs dramatically
- Agentic coding: Antigravity + Gemini CLI, with SWE-bench 78% even on the Flash tier
- Multimodal: native image/video/audio input, plus the Nano Banana Pro image-generation tier
Limitations
- The whole family is still in preview; rate limits and model IDs may change
- Prompts beyond 200K tokens double in price ($4/$18)
- Knowledge cutoff is January 2025 - rely on grounding for current information
Frequently asked questions
What does Gemini 3 Pro cost?
Official pricing: $2 input / $12 output per million tokens for prompts up to 200K tokens; $4/$18 beyond. It is the cheapest frontier flagship (ai.google.dev/pricing).
Is Gemini 3 free?
AI Studio offers a rate-limited free tier; the API is usage-based, with Batch at half price and caching saving up to 90% on input.
Pro or Flash?
Choose 3.1 Pro for the strongest reasoning; choose 3 Flash for near-equal intelligence at 3x the speed and a quarter of the price (GPQA 90.4%, SWE-bench 78% is already strong).
What is the latest version?
As of October 2026: gemini-3.1-pro-preview (1M context) on the Pro line, and gemini-3-flash-preview / gemini-3.1-flash-lite on the Flash line.
*This page is not affiliated with Google. Pricing verified as of October 2026 (ai.google.dev/pricing); confirm current figures on the official site.*