← AI Tools directory
General AssistantResearch dossier

Gemini

Google's multimodal assistant across the Gemini app, Search, Workspace, Android and developer platforms.

Gemini is both a consumer product and a model family. Gemini 4 Argon is Google's new flagship sparse MoE reasoning model, while Gemini 3.8 Flash handles high-throughput and coding workloads.

First releasedFebruary 8, 2024 (Gemini brand; Bard launched in 2023)
CompanyGoogle
Best forGoogle ecosystem users, multimodal work, Search and Workspace integration
PricingFree; Google AI Plus/Pro/Ultra pricing varies by market
01 / OVERVIEW

What Gemini is today

Google renamed Bard to Gemini in February 2024 and paired the app with its Gemini model family, mobile access and Google AI subscriptions.

The product combines chat with Google Search grounding, connected apps, long context, live multimodal interaction, Deep Research and Gemini 4 Argon frontier reasoning.

Reading note

Product features and underlying models are documented separately. Roadmap items include only provider-announced work; rumors are excluded.

02 / RELEASE HISTORY

Major versions and what changed

Selected milestones that materially changed the product.

5 milestones
Current

Gemini 4 Argon frontier release

Google rolled out Gemini 4 Argon, a large sparse MoE transformer for deep multi-step reasoning, coding, science and robotics.

  • Tunable thinking levels: low, medium, high
  • 1M unified input-output context window
  • Introductory API pricing: $2 input / $10 output per 1M tokens
  • Rolled out to Google AI Pro/Ultra subscribers and Vertex AI
Official release note
Milestone

Bard becomes Gemini

Google introduced the Gemini name, Gemini Advanced with Ultra 1.0 and dedicated mobile access.

  • Ultra 1.0 access
  • Android app and iOS entry point
  • Text, voice and image prompts
  • Expanded Workspace integration roadmap
Official release note
Milestone

Gemini 2.0

The family moved toward agentic systems with native tool use and multimodal output prototypes.

  • Native Search and Maps tool use
  • Image and audio output research
  • Project Astra, Mariner and Jules prototypes
  • 2.0 Flash preview
Official release note
Milestone

Gemini 2.5

Thinking models, 1M-token context, stronger coding and Deep Think expanded complex work.

  • Thinking models
  • 1M-token context
  • Deep Research upgrades
  • Native audio and computer-use previews
Official release note
Milestone

Gemini 3.1 Pro and 3.5 Flash

3.1 Pro more than doubled 3 Pro's ARC-AGI-2 result; 3.5 Flash then targeted frontier quality at much higher speed.

  • 77.1% ARC-AGI-2 for 3.1 Pro
  • Stronger visual and interactive coding
  • 3.5 Flash outperforms 3.1 Pro across almost all cited benchmarks
  • Four-times-faster provider claim for agent workflows
Official release note
03 / THEN VS NOW

Gemini 3 Pro vs 3.1 Pro, then 3.5 Flash

The comparison uses provider-published capabilities and named evaluations, not a generic marketing score.

DimensionPreviousCurrentMeasured difference
Novel reasoningGemini 3 Pro baseline3.1 Pro: 77.1% ARC-AGI-2More than 2x the 3 Pro score, according to Google
Speed3.1 Pro optimized for difficult reasoning3.5 Flash optimized for frontier speedGoogle says 3.5 Flash runs 4x faster than other frontier models
Product workflowChat plus connected Google appsMore agentic coding, Deep Research and live multimodal workflowsMoves from answer generation toward planning and execution
Model choicePro/Flash split3.1 Pro depth plus 3.5 Flash speedBest model depends on reasoning depth, latency and product surface
04 / POWER EVIDENCE

One official result is genuinely above 2x - but only on a specific reasoning benchmark.

Google reports 3.1 Pro scored more than twice Gemini 3 Pro on ARC-AGI-2. That does not mean every response or feature is twice as capable. Separately, Google describes 3.5 Flash as four times faster than other frontier models, which is a speed claim, not 4x intelligence.

>2x

ARC-AGI-2 score

Gemini 3.1 Pro reached a verified 77.1%, more than double Gemini 3 Pro's result.

Check evidence
4x

Claimed model speed

Google says Gemini 3.5 Flash runs four times faster than other frontier models while beating 3.1 Pro on almost all cited benchmarks.

Check evidence
1M

2.5 Pro context

The 2.5 generation established million-token long-context workflows in supported products.

Check evidence
05 / FEATURES EXPLAINED

How the important features work

What happens behind the button, where it helps and what to verify.

01

Deep Research

Limits and model vary by plan

How it works: Builds a research plan, searches the web and connected sources, then synthesizes a linked report.

Best use: Complex comparisons, background research and decision briefs

Watch: Inspect citations and dates; synthesis can omit conflicting evidence.

02

Gemini Live

Device, language and plan dependent

How it works: Streams audio and, on supported devices, camera or screen context for a continuous spoken conversation.

Best use: Hands-free tutoring, visual assistance and brainstorming

03

Connected Google apps

Account and Workspace controls apply

How it works: Extensions retrieve permitted context from services such as Gmail, Drive, Maps or YouTube and pass relevant information into the response.

Best use: Email summaries, travel planning and document workflows

Watch: Review which services are connected and what organizational policies allow.

04

Canvas and creation tools

Feature availability varies

How it works: Creates editable documents, code, quizzes and media-oriented outputs in a side-by-side workspace.

Best use: Drafting, learning materials and rapid prototypes

LIMITS

What the headline features do not guarantee

  • The Gemini app, Gemini models, Workspace features and API do not share identical model access or limits.
  • Live web grounding improves freshness but does not guarantee correct source selection or interpretation.
  • Preview model names, pricing and limits can change before general availability.
06 / WHAT'S NEXT

Announced roadmap, without speculation

Status reflects the provider's latest public wording. “No public date” means there is no verified launch schedule.

Rolling out

Gemini 3.5 Flash expansion

Google announced broader agentic development use following the I/O 2026 release.

Official announcement
Announced

3.1 Pro general availability

Google said the preview would advance agentic workflows before general availability, but the launch post did not give a fixed date.

Official announcement
No public date

Next numbered generation

No official release date or complete specification for Gemini 4 has been published.

07 / SOURCES

Primary sources

Provider announcements, help documentation and system information.

3 references
  1. 1
    Bard becomes Gemini

    Google · Feb 8, 2024

  2. 2
    Gemini 3.1 Pro

    Google · Feb 19, 2026

  3. 3

Last editorial verification: 2026-10-15. Prices, access limits, preview status and model routing can change after this date.