← Models

Gemini 3.1 Pro Preview

Google · Gemini 3.1

Google's current premium Gemini 3.1 preview model for multimodal reasoning, coding, long-context analysis, and agentic workflows.

Part of Gemini Pro family · Other versions: Gemini 2.5 Pro TTS Preview , Gemini 3 Pro Preview , Computer Use Preview , Gemini 2.5 Pro
Type
multimodal
Context
1M tokens
Max Output
66K tokens
Status
preview
Input
$2/1M tok
Output
$12/1M tok
API Access
Yes
License
proprietary
multimodal long-context reasoning coding vision audio preview
Released March 2026 · Updated May 16, 2026

Overview

Freshness note: Model capabilities, limits, and pricing can change quickly. This profile is a point-in-time snapshot last verified on May 16, 2026.

Gemini 3.1 Pro Preview is Google’s current premium Gemini preview model, positioned as the migration target for the shut down Gemini 3 Pro Preview. Google’s May 2026 Gemini API docs list it at the top of the Gemini 3 section alongside Gemini 3 Flash, Gemini 3.1 Flash-Lite, Gemini 3.1 Flash Live, and Gemini 3.1 Flash TTS. It is aimed at multimodal understanding, agentic capabilities, coding-heavy workflows, and long-context reasoning where teams want Google’s newest premium model behavior before a stable release arrives.

Unlike Gemini 2.5 Pro, Gemini 3.1 Pro Preview is still a preview model. That means behavior, pricing, and lifecycle guarantees can change faster than on stable Gemini 2.5 tiers, so production teams should use stronger evaluation and fallback discipline.

Capabilities

Gemini 3.1 Pro Preview is one of Google’s broadest capability models:

  • Native multimodal understanding across text, images, audio, video, and PDFs
  • 1M-token context enables very large codebase analysis, long document review, and broad multimodal evidence synthesis in a single pass.
  • Strong reasoning and coding performance, with Google publicly highlighting improved benchmark performance over Gemini 3 Pro.
  • Grounded generation with the ability to cite sources and integrate with Google Search for factual accuracy.
  • Structured output support for JSON, function calling, file search, and tool use in agentic workflows.

The practical distinction is not just raw scale beyond stable Gemini 2.5 Pro, but access to Google’s current premium preview lane for harder multimodal and agentic tasks.

Technical Details

Google’s current models and pricing docs position Gemini 3.1 Pro Preview as the newest premium preview model and list both current model IDs:

  • gemini-3.1-pro-preview
  • gemini-3.1-pro-preview-customtools

The same docs publish:

  • 1,048,576 token input window
  • 65,536 max output tokens
  • Text, image, audio, video, and PDF input
  • Text and structured outputs
  • Built-in grounding with Google Search and Google Maps
  • Code execution, file search, function calling, caching, and URL context

The model is available through Google AI Studio and Vertex AI deployment paths where preview models are enabled. The current all-models page also lists Gemini Deep Research, Computer Use Preview, Nano Banana 2, Nano Banana Pro, Veo 3.1, Lyria 3, and Gemini Embedding 2 alongside the Gemini 3 family, which is useful context when choosing adjacent Google models.

Pricing & Access

Current API pricing (per 1M tokens, for prompts up to 200K tokens):

  • Input: $2.00
  • Output: $12.00

Pricing rises to 4/4 / 18 above 200K-token prompts, and preview lifecycle risk is higher than on Gemini 2.5 Pro. Batch pricing and context caching are also published, which matters for retrieval-heavy or large-prefix workflows.

Access options:

  • Google AI Studio
  • Vertex AI deployment paths
  • Gemini app and NotebookLM on higher Google AI plans
  • Google Gemini API integrations that allow preview-model usage

Google’s current docs also position Gemini 3.1 Pro across the Gemini API and Google Cloud deployment paths. For teams already on Google Cloud, Vertex AI remains the most controlled route for preview adoption and governance.

Best Use Cases

Gemini 3.1 Pro Preview is strongest in scenarios that benefit from Google’s newest preview-tier multimodal reasoning:

  • Whole-codebase analysis and refactoring where fitting a large repo slice in context reduces chunking complexity.
  • Long document review — contracts, research papers, regulatory filings — in a single prompt.
  • Multimodal workflows combining images, audio, or video with text analysis.
  • Research and synthesis across large collections of sources.
  • Applications that benefit from grounded, citation-backed responses via Search integration.

Less ideal for latency-sensitive, short-context tasks where Gemini 2.5 Flash or Flash-Lite deliver comparable quality at lower cost and faster speeds.

Comparisons

  • Gemini 2.5 Pro (Google): Safer stable route for premium reasoning, while 3.1 Pro is the newer preview lane.
  • Claude Opus 4.6 (Anthropic): Opus is a strong alternative for instruction-heavy enterprise reasoning; Gemini tends to differentiate on first-party multimodal breadth.
  • GPT-5.5 (OpenAI): Competitive on reasoning and coding, with stronger OpenAI-native tooling around ChatGPT and Codex.