Gemini 3

What is Gemini 3?

Gemini 3 is not one model but Google's whole current line-up on the Gemini API and in the Gemini app. The text models read text, images, video, audio and PDFs across a 1M-token context and write up to 64K tokens. Around them sit Live voice models, text-to-speech, transcription and the Nano Banana image models, and most of the family is free to try on a rate-limited tier.

Google ships new versions every few weeks. 3.8 Flash is its strongest Flash model, while 3.7 Flash and 3.6 Flash stay supported. The original Gemini 3 Pro is shut down; 3.1 Pro replaces it but is still a preview.

Which Gemini 3 models are current?

Google's current picks are 3.8 Flash for most work, 3.1 Flash-Lite for cheap bulk jobs and the 3.1 Pro preview for the hardest reasoning; the original Gemini 3 Pro is shut down.

Gemini 3.8 Flash

The newest and strongest Flash model, stable, and Google's default for coding and agents.

Gemini 3.7 Flash and 3.6 Flash

Earlier Flash models, still supported at the same price; Google suggests 3.7 Flash when token efficiency matters most.

Gemini 3.1 Pro

Still a preview; it replaced the shut-down Gemini 3 Pro Preview.

Gemini 3.1 Flash-Lite

Stable, but Google has set a shutdown date and names Gemini 3.5 Flash-Lite as its replacement.

Gemini 3 Flash

A preview that Google now calls legacy, with Gemini 3.6 Flash as the recommended replacement.

Who is Gemini 3 best for?

Gemini 3 is best for developers who want fast, low-cost models with a free tier and Google Search built in, from bulk jobs on 3.1 Flash-Lite to coding and agents on 3.8 Flash.

• Fast coding and prototyping

  Gemini 3.8 Flash is Google's strongest Flash model for software and agent work, with a 1M-token context at $1.50 in and $7.50 out per million tokens.

• High-volume, low-cost jobs

  Gemini 3.1 Flash-Lite handles translation, tagging and extraction at $0.25 in and $1.50 out per million tokens.

• Answers that need fresh facts

  Grounding with Google Search and Google Maps is built in, with 5,000 free requests a month on the paid tier, then $14 per 1,000.

• Voice and speech apps

  Live models hold real-time spoken conversations, and separate models handle text-to-speech in up to 130 languages and transcription with speaker labels.

What's included with Gemini 3?

A free API tier, adjustable thinking levels, half-price Batch and Flex requests, Google Search and Maps grounding, built-in tools and context caching.

Free API tier

Most Flash, Flash-Lite, Live and speech models are free at lower rate limits, though free-tier prompts are used to improve Google's products.

Thinking levels

You set how much a model reasons before it answers, from minimal or low up to high; 3.8 Flash defaults to medium and 3.1 Pro to high.

Batch, Flex and Priority

Batch and Flex requests cost half the Standard rate for work that can wait; Priority costs more, for example $2.70 / $13.50 on 3.8 Flash.

Search and Maps grounding

Models can check Google Search and Google Maps before answering, which helps with anything newer than their training data.

Built-in tools

Code execution, URL context, file search, function calling and structured JSON output work across the text models.

Context caching

Cached input costs a tenth of the normal input price, plus an hourly storage fee, which cuts the bill for long repeated prompts.

How much does Gemini 3 cost?

Gemini 3 is free on the Gemini API's rate-limited tier for most models; paid text prices run from $0.25 in and $1.50 out per million tokens for 3.1 Flash-Lite to $2.00 in and $12.00 out for 3.1 Pro Preview.

Other paid prices per million input / output tokens: Gemini 3.5 Flash $1.50 / $9.00, Gemini 3.5 Flash-Lite $0.30 / $2.50 and Gemini 3 Flash Preview $0.50 / $3.00. The Live voice models charge $3.00 per million audio input tokens and $12.00 per million audio output tokens, and Nano Banana 2 images cost about $0.045 to $0.151 each depending on size. Google Search grounding gives 5,000 free requests a month across Gemini 3 models, then $14 per 1,000. The paid tier needs a linked billing account and at least $5 of prepaid credit, which expires after a year, and paid prompts are not used to improve Google's products. In the Gemini app, the Free plan includes 3.6 Flash and some 3.1 Pro use, and Google AI Pro costs $19.99 a month (US prices). Enterprise use on Gemini Enterprise Agent Platform is priced through sales. Full pricing on ai.google.dev. We checked these prices on 28 Sep 2026.

How do readers rate Gemini 3?

No one has rated Gemini 3 yet. Be the first.

Gemini 3 pros and cons

It is fast, cheap and grounded in Google Search, but developers find it weaker in long agentic coding runs, and its Pro model is still a preview.

What are the best Gemini 3 alternatives?

Three tools on Recaply do Gemini 3's job. Gemma 4 is open source, so you can run it on your own server.

Claude Sonnet 5Free

Choose it when agentic coding matters most and you want a stable model at a similar price: $2 / $10 per million tokens with a 1M context, and its knowledge stops at January 2026.

GPT-6 AstraPaid

Choose it when you need OpenAI's most capable model for the hardest end-to-end work; it costs $10 / $50 per million tokens, five times Gemini 3.1 Pro's input price, and its knowledge stops at April 2026.

Gemma 4Open Source

Choose it when you need to run the model on your own hardware or offline; Google publishes its weights under Apache 2.0 in five sizes, from the phone-sized E2B to 31B, with up to a 256K context.

Questions about Gemini 3