---
title: "Gemini 3.1 Pro"
type: model
id: "gemini-3.1-pro"
provider: "Google"
model_type: "proprietary"
release_date: "2026-02"
description: "Google's most capable model, building on the Gemini 3 series with improved reasoning, multimodal, and agentic capabilities. Native multimodal across text, images, video, and audio."
last_updated: "2026-04-10"
context_window: "1M tokens"
website: "https://deepmind.google"
license: "Proprietary"
modality:
- "text"
- "image"
- "video"
- "audio"
tags:
- "google"
- "proprietary"
- "text"
- "image"
- "video"
- "audio"
pricing:
  input: "$2.00 / 1M tokens"
  output: "$12.00 / 1M tokens"
  note: "Free tier via Google AI Studio"
benchmarks:
  reasoning: 93
  coding: 91
  math: 92
  writing: 89
  multilingual: 93
  speed: 78
best_for:
- "Long document analysis"
- "Multimodal tasks"
- "Google ecosystem integration"
- "Research"
---

# Gemini 3.1 Pro

The best value among frontier models. At $2/$12 per million tokens, Gemini 3.1 Pro delivers 93/100 reasoning and 92/100 math -- numbers that rival GPT-5.4 at less than half the price. Add a generous free tier through Google AI Studio and this becomes the easiest model to start experimenting with.

The multimodal story is where Gemini genuinely leads. It's the only frontier model with native support across text, images, video, and audio -- not bolted-on, but trained natively. If your workflow involves analyzing video content, processing audio, or mixing modalities, no competitor matches this breadth. The 93/100 multilingual score also makes it the strongest proprietary option for non-English work alongside Qwen 3.

The weakness is consistency. Gemini can be brilliant on one prompt and mediocre on a slight rephrasing. Claude and GPT tend to be more predictable. The coding score of 91/100 is strong but trails both Opus 4.6 (97) and Sonnet 4.6 (93). Writing at 89/100 is competent but clearly behind the Claude models.

**When to pick something else:** For coding-heavy work, Claude Sonnet 4.6 or Opus 4.6 is the better bet. For maximum reasoning ceiling, GPT-5.4 or GPT-5.4 Thinking pulls ahead. But if you want a strong generalist at an aggressive price point with the best multimodal support available, Gemini 3.1 Pro is the move.