Skip to main content
Skip to main content

Gemini 3 Pro Preview

Gemini 3 Pro Preview was Google's first Gemini 3 series model, released November 2025 - a frontier reasoning and multimodal understanding model with strong agentic and coding capabilities.

  • 99.9% SLA
  • Official Discounts
  • Pay-as-you-go
  • Fast & Low Latency
Brand:GoogleInput token: $1.80/MOutput token: $10.80/MInput cache read: $0.18/MInput cache write: $0.3375/M

Gemini 3 Pro Preview: Google's First Gemini 3 Frontier Model - Full Review

Gemini 3 Pro Preview (this catalog entry: gemini-3-pro-preview) was Google's first Gemini 3 series model, released November 2025 - a frontier reasoning and multimodal understanding model with strong agentic and coding capabilities.

Here's the short version: Gemini 3 Pro Preview opened the Gemini 3 generation - 1M-token context, full multimodal input (text, image, video, audio, PDF), and Google's frontier reasoning stack aimed at agents and coding. It set the pattern that Gemini 3.1 Pro Preview then refined (better thinking, higher token efficiency, more grounded and consistent). The honest reality in 2026: the preview model has been shut down per the Gemini API docs - production work belongs on the current Gemini 3.x line.

This guide covers model overview, core features, technical specifications, capability comparison, core advantages, recommended use cases, example prompts, selection recommendations, status, and my verdict.

Quick Facts

AttributeValue
Catalog sluggemini-3-pro-preview
Model IDgemini-3-pro-preview
DeveloperGoogle DeepMind
ReleaseNovember 2025
PositionFirst Gemini 3 series model
Context1,048,576 input tokens
InputsText, image, video, audio, PDF
FocusReasoning, agentic, coding
StatusShut down (per Gemini API docs)
SuccessorGemini 3.1 Pro Preview

Table of Contents

  1. Model Overview
  2. Core Features
  3. Technical Specifications
  4. Capability Comparison
  5. Core Advantages
  6. Recommended Use Cases
  7. Example Prompts
  8. Selection Recommendations
  9. Status: Preview, Then Shutdown
  10. The Bottom Line
  11. FAQ
  12. Sources & Further Reading

1. Model Overview

Gemini 3 Pro Preview launched in November 2025 as the first model of the Gemini 3 series - Google's statement that the next frontier was reasoning and multimodal understanding for agents and coding. It accepted text, image, video, audio, and PDF inputs within a 1M-token context.

The model's launch followed the Gemini 3 Pro Image preview and established the pattern for the series: frontier reasoning, deep multimodal understanding, and agentic tooling. Gemini 3.1 Pro Preview (the next release) then improved performance and stability with better thinking, higher token efficiency, and more grounded, consistent responses.

For this catalog entry, the review target is the preview itself - with the honest operational reality that the model has been shut down per the Gemini API docs, and current work belongs on the active Gemini 3.x models.

2. Core Features

Frontier reasoning. Google's top-tier reasoning stack for complex tasks.

Multimodal understanding. Text, image, video, audio, and PDF input.

Agentic capabilities. Tool use and multi-step workflows.

Coding strength. The launch focus alongside agents.

1M context. Full-document and long-conversation input.

Gemini 3 series foundation. The template for 3.1 and successors.

3. Technical Specifications

SpecificationDetail
Modelgemini-3-pro-preview
Context1,048,576 input tokens
InputsText, image, video, audio, PDF
OutputText (+ tool calls)
ReleaseNovember 2025
StatusShut down (per Gemini API docs)
SuccessorGemini 3.1 Pro Preview

Warning: The Gemini API model page lists gemini-3-pro-preview as shut down. Do not build new integrations against it.

4. Capability Comparison

CapabilityGemini 3 Pro PreviewGemini 3.1 Pro PreviewGemini 3.7 FlashGemini 2.5 Pro
ReasoningFrontier (launch)ImprovedWorkhorseStrong
Context1M1M1M1M
Token efficiencyBaselineHigherImprovedBaseline
StatusShut downCurrentCurrentDeprecated
Best forHistoricalCurrent frontierProduction workhorseLegacy

Reading the table honestly: 3 Pro Preview was the launch vehicle; 3.1 Pro Preview is its refined form; 3.7 Flash is the production workhorse. Choose by status, not nostalgia.

5. Core Advantages

  1. Frontier reasoning at launch. The Gemini 3 series' capability statement.
  2. Full multimodal input. Text, image, video, audio, PDF.
  3. Agentic + coding focus. The workloads Google prioritized.
  4. 1M context. Document-scale reasoning.
  5. Series foundation. The pattern successors refined.

6. Recommended Use Cases

Historical (Nov 2025 - shutdown):

  • Frontier reasoning benchmarks.
  • Agentic and coding evaluation.
  • Multimodal understanding testing.

Current (2026):

  • None - the model is shut down. Use the active Gemini 3.x models.

7. Example Prompts

1. Reasoning

text
Analyze this multi-file codebase for a race condition: trace the request
path, identify the root cause, and propose a fix with a regression test.

2. Multimodal

text
Given this PDF contract, this spreadsheet, and this video transcript,
summarize the obligations and flag conflicts.

3. Agentic

text
You have tools: search_kb, get_order_status, refund_order. A customer
reports a missing order. Verify, check the delivery log, draft a response.

Prompting guidance (historical):

  • The model followed Gemini 3 family prompting: explicit reasoning requests, tool descriptions, and structured output.
  • For current work, use the active models' prompting guides.

8. Selection Recommendations

Do not select gemini-3-pro-preview:

  • It is shut down. No integration, no production use.

Choose Gemini 3.1 Pro Preview (or current Pro line) if:

  • You need the frontier reasoning tier with improved efficiency.

Choose Gemini 3.7 Flash if:

  • You need the production workhorse with agentic scores at Flash pricing.

9. Status: Preview, Then Shutdown

Rendering flowchart...

Practical notes:

  • Any tutorial referencing gemini-3-pro-preview is obsolete.
  • Migrate workloads to the current Gemini 3.x models.
  • The model's value now is historical and architectural.

10. The Bottom Line

Verdict: A landmark preview, now retired - migrate to the active line. Gemini 3 Pro Preview opened the Gemini 3 generation with frontier reasoning, full multimodal input, and a 1M context - and its architecture shaped everything that followed. But the model is shut down per the API docs, and production work belongs on Gemini 3.1 Pro or the 3.7 Flash workhorse. Treat this entry as history; build on the current models.

Sources & Further Reading

Status reflects Google's official Gemini API documentation as of August 2026. The preview model is shut down; verify current models before integration.