AI/TLDR

Gemini 3.1 Pro

Google's flagship reasoning model for the most complex agentic and coding tasks.

Overview

Gemini 3.1 Pro is Google DeepMind's flagship model, released February 19, 2026 as the latest member of the Gemini 3 family. It is positioned for tasks where a simple answer isn't enough, with advanced reasoning, improved agentic capabilities, and strong multimodal understanding across text, audio, images, video, and entire code repositories.

The model accepts up to a 1M-token input context window and produces up to 64K output tokens, making it suited to long-context understanding over large documents and codebases. Knowledge cutoff is January 2025. It is available in preview via the Gemini API on Google AI Studio, Gemini CLI, Vertex AI, Gemini Enterprise, and the Gemini app for Google AI Pro/Ultra users.

On Google's published evaluations, Gemini 3.1 Pro reaches 94.3% on GPQA Diamond, 77.1% on ARC-AGI-2 (described as more than double the reasoning performance of Gemini 3 Pro), and 80.6% on SWE-Bench Verified, placing it among the strongest models on reasoning, agentic, and coding tasks.

Released2026-02-19
LicenseProprietary
WeightsAPI only
ParametersUndisclosed
Context1M
Max output64K
ArchitectureUndisclosed (Gemini 3 family)
Knowledge cutoffJanuary 2025
ModalitiesText, Vision, Audio, Video, PDF
StatusPreview

Benchmarks

  1. GPQA Diamond94.3%
  2. MMMLU (Multilingual Q&A)92.6%
  3. ARC-AGI-277.1%
  4. SWE-Bench Verified80.6%
  5. MRCR v2 (128k)84.9%
  6. Humanity's Last Exam (no tools)44.4%
  7. Terminal-Bench 2.068.5%
  8. SWE-Bench Pro (Public)54.2%
  9. SciCode59%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

Pricing

Input$2.00 / 1M tokens
Output$12.00 / 1M tokens

Standard paid tier for prompts <=200k tokens; $4.00 input / $18.00 output for prompts >200k tokens.

Pricing source ↗

Strengths

  • Frontier-level reasoning, with 94.3% on GPQA Diamond and 77.1% on ARC-AGI-2
  • Strong agentic and coding performance (80.6% SWE-Bench Verified, 68.5% Terminal-Bench 2.0)
  • 1M-token input context for long-document and whole-repository understanding
  • Native multimodal input across text, image, audio, video, and PDF
  • Strong long-context retrieval (84.9% on MRCR v2 at 128k)

Best for

  • Agentic workflows and multi-step tool use
  • Advanced software engineering and codebase-scale coding
  • Long-context analysis of large documents and repositories
  • Complex multimodal reasoning over text, images, audio, and video

How to access

ProviderModel ID
Google AI / Gemini API ↗gemini-3.1-pro-preview

Gemini Pro — every version

The full lineage of the Gemini Pro line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.

VersionReleasedContextLicense
Gemini 3.5 Pro2026-05-19Proprietary
Gemini 3.1 Procurrent2026-02-191MProprietary
Gemini 3 Pro2025-11-18Proprietary
Gemini 2.5 Pro2025-03-25Proprietary
Gemini 2.0 Pro2025-02-05Proprietary
Gemini 1.5 Pro2024-02-15Proprietary
Gemini 1.0 Ultra2024-02-08Proprietary
Gemini 1.0 Pro2023-12-13Proprietary

FAQ

What is Gemini 3.1 Pro?

Gemini 3.1 Pro is Google DeepMind's flagship model, released February 19, 2026 as part of the Gemini 3 family. It targets the most complex reasoning, agentic, and coding tasks, accepts multimodal input across text, image, audio, video, and PDF, and offers a 1M-token input context window with up to 64K output tokens.

How much does the Gemini 3.1 Pro API cost?

On the standard paid tier, Gemini 3.1 Pro Preview costs $2.00 per 1M input tokens and $12.00 per 1M output tokens for prompts up to 200k tokens. For prompts over 200k tokens, pricing rises to $4.00 input and $18.00 output per 1M tokens. Output pricing includes thinking tokens.

What is the Gemini 3.1 Pro context window?

Gemini 3.1 Pro supports an input context window of up to 1 million tokens and can produce up to 64K output tokens. The large context enables long-document analysis and reasoning over entire code repositories. Its training knowledge cutoff is January 2025.

Are Gemini 3.1 Pro weights open?

No. Gemini 3.1 Pro is a proprietary, closed-weight model. It is available only through Google's services, including the Gemini API on Google AI Studio, Gemini CLI, Vertex AI, Gemini Enterprise, and the Gemini app for Google AI Pro and Ultra subscribers. There are no downloadable weights.