Gemini 3.1 Pro Model Card
by Google DeepMind
The official spec for Google's frontier Gemini 3.1 Pro: context window, modalities, benchmarks, and agentic strengths.
Overview
The Gemini 3.1 Pro Model Card, published by Google DeepMind on February 19, 2026, is the official specification for Gemini 3.1 Pro, a model built on Gemini 3 Pro in the natively multimodal, mixture-of-experts Gemini reasoning family. It documents an input context window of up to 1 million tokens across text, images, audio, video, and entire code repositories, with text output up to 64K tokens, and describes a Deep Think mode for harder problems. The card reports that Gemini 3.1 Pro significantly outperforms Gemini 3 Pro on reasoning- and multimodal-heavy benchmarks, including 77.1% on ARC-AGI-2 (abstract reasoning) and 94.3% on GPQA Diamond (graduate-level science), and highlights strengths in agentic performance, advanced coding, long-context and multimodal understanding, and algorithmic development. As a model card it also covers intended uses, safety framing, and evaluation methodology — the primary-source facts an AI engineer should read before adopting the model.
At a Glance
- Topic
- Models
- Level
- Advanced
- Format
- Documentation
- Cost
- Free
- Duration
- ~20 min read
- Provider
- Google DeepMind
- Hands-on
- No
- Certificate
- None
What You’ll Learn
- ✓Gemini 3.1 Pro's context window (up to 1M input tokens) and supported modalities
- ✓Reported benchmark results, including ARC-AGI-2 (77.1%) and GPQA Diamond (94.3%)
- ✓Where the model claims strength: agentic performance, advanced coding, long-context and multimodal reasoning
- ✓What Deep Think mode is and when to reach for it
- ✓How Google DeepMind frames intended uses, limitations, and evaluation for a frontier model
Highlights
- •Authoritative primary source straight from Google DeepMind
- •One of the newest frontier-model references (published February 2026)
- •Concrete numbers for context window, output limits, and headline benchmarks
Who It’s For
Best For
- ✓AI engineers evaluating Gemini 3.1 Pro for agentic, coding, or multimodal workloads
- ✓Practitioners comparing frontier model capabilities and context limits
- ✓Anyone who wants primary-source facts instead of secondhand benchmark claims
Prerequisites
- •Familiarity with LLM capabilities and benchmarks
- •Understanding of context windows and multimodal inputs
FAQ
What is Gemini 3.1 Pro Model Card?
The official Google DeepMind model card for Gemini 3.1 Pro, a natively multimodal frontier reasoning model. It's the authoritative reference for engineers deciding whether and how to build on Gemini 3.1 Pro, documenting its capabilities, limits, and evaluation results.
Is Gemini 3.1 Pro Model Card free?
Gemini 3.1 Pro Model Card is free to access.
What level is Gemini 3.1 Pro Model Card for?
Gemini 3.1 Pro Model Card is aimed at a advanced audience. Recommended background: Familiarity with LLM capabilities and benchmarks, Understanding of context windows and multimodal inputs.
How long does Gemini 3.1 Pro Model Card take?
Expect roughly ~20 min read. Most learners work through it at their own pace.
What will I learn from Gemini 3.1 Pro Model Card?
You'll learn: Gemini 3.1 Pro's context window (up to 1M input tokens) and supported modalities; Reported benchmark results, including ARC-AGI-2 (77.1%) and GPQA Diamond (94.3%); Where the model claims strength: agentic performance, advanced coding, long-context and multimodal reasoning; What Deep Think mode is and when to reach for it; How Google DeepMind frames intended uses, limitations, and evaluation for a frontier model.