Skip to content

Google · Fast work with documents, images and audio

Gemini 3.6 Flash

A fast model for documents, images, video and audio. It is suitable for a first pass that finds and organizes information from a large volume of source material.

Compare with another model

Costs
Medium · $1.50 per million input tokens and $7.50 per million output tokens; a token is a small piece of text
Speed
Very fast · generated 304 tokens per second in the measurement
Availability
Online service from Google and Gemini products
Input length
Very long input · up to 1 million tokens (small pieces of text)
Inputs
Text, images, video and audio as input

Data checked . Release date unverified.

Indicative capability profile

Where the model is strong

The five levels are our clear summary of the results below. They are not a ranking that applies to every task.

Analysis Solves practical problems well, but does not reach the very top tier.
Coding Usable code, but not the first choice for the most difficult programs.
Autonomous work across multiple steps A good choice for quick work with documents, images and audio.
Writing and content Suitable for finding data and summarizing; check important outputs.
Speed 304 output tokens per second in the test.

Ideal use

When to choose it

  • Finding and sorting information in documents and media
  • Interactive assistants with fast responses
  • First pass through long documents

Usage boundaries

When to choose another model

  • Final strategic decision without further review
  • The hardest scientific and logic problems

Traceable supporting sources

Data and measurement sources

Distinguish between the manufacturer's documentation and the results of a specific test. Measurements also depend on the settings and the task set used.

Artificial Analysis

Intelligence Index v4.1

50 points

Performance close to the best models in significantly less time.

Open original source ↗
Artificial Analysis

Time and cost per task

1.3 min · $0.50

A practical signal for fast, high-volume processing.

Open original source ↗

Trends over time

Related events from AI Radar

Search more →
Google important update

DeepMind changes leadership: Demis Hassabis steps away from operations, replaced by Koray Kavukcuoglu, following the departure of the founders of the startup Discovery Loop

New: Demis Hassabis is stepping away from operations at DeepMind; Koray Kavukcuoglu is becoming the new leader of DeepMind and the remaining co-leader of Gemini; More senior talent in reinforcement learning is leaving DeepMind

For people in AI/ML, the departure of such a prominent group of researchers confirms that founding independent research startups (known as neolabs) funded by venture capital is now a viable alternative to a career at major laboratories such as DeepMind.

The leadership restructuring at DeepMind, together with the departure of four key researchers to a newly funded competitor, increases the risk of further talent departures from Google; at the same time, Google is retaining a stake in the new startup and becoming its cloud partner, so it retains some of the value and future research results even after the researchers leave.

✓ 3
GitHub important update

GitHub recommends Claude Opus 5.5 following the retirement of Claude Opus 4.7

New: The recommended replacement for Claude Opus 4.7 is Claude Opus 5.5, not Claude Opus 5.

When switching from Claude Opus 4.7, follow the current recommendation to use Claude Opus 5.5; the earlier announcement listed a different replacement.

Migrating company workflows and integrations to supported models may require enabling replacement models through access policies in Copilot Enterprise or Copilot Business.

✓ official
Google important

Requests to the model gemini-3.5-flash are automatically redirected to the model gemini-3.6-flash

A developer using Gemini API has two recommended options when updating the configuration: the identifier gemini-3.6-flash or gemini-3.8-flash.

Model management in business applications must account for the automatic change: the original identifier in the deployed configuration no longer refers to the model that actually processes requests.

✓ official
Google important

Google released the code for Glanceboard, a home application with AI illustrations on an e-ink display

Anyone interested in home AI projects can download the open-source code and use the included guide and hardware list to build their own device that displays a schedule and recommended clothing based on the weather on an e-ink display without the cloud.

The project serves as a public example of how Google combines Gemini 3.6 Flash and Nano Banana models with Google Antigravity to quickly build a "vibe-coded" local application connected to hardware, which may inspire teams looking for an example to follow for similar rapid prototypes without relying on the cloud.

✓ official

Next step

Compare the model using your actual task.

A benchmark narrows the selection. A short trial on your data determines the choice.

Open comparison →