Benchmark-backed reasoning gains
ARC-AGI-2 and related benchmark results indicate stronger capability for unfamiliar logic and complex problem solving.
Gemini 3.1 Pro
Google's most advanced Pro-tier AI model, delivering breakthrough reasoning with 77.1% on ARC-AGI-2 — more than double its predecessor. Built for your hardest challenges with a 1M token context window and native multimodal understanding.
ARC-AGI-2 Score
77.1%
Context Window
1M tokens
Max Output
64K tokens
Humanity's Last Exam
44.4%
Try It Now
Try Gemini 3.1 Pro
Experience Google's most advanced reasoning model — enter a prompt and see breakthrough intelligence in action.
Try a sample prompt
5 credits / message
Sign in to try
Core Capabilities
What Gemini 3.1 Pro Can Do
Advanced multimodal reasoning engineered for complex, real-world tasks that demand more than a simple answer.
Gemini 3.1 Pro
SOTA Performance
MMLU92%
HumanEval87%
MATH79%
01
Advanced Reasoning
Scores 77.1% on ARC-AGI-2 — more than double Gemini 3 Pro — solving entirely new logic patterns and complex problem-solving tasks.
02
Multimodal Understanding
Natively processes text, images, audio, video, and code in a single context, enabling rich cross-modal reasoning and synthesis.
03
Agentic Workflows
Optimized for ambitious agentic tasks via Google Antigravity, Gemini CLI, and Vertex AI — orchestrating multi-step workflows at scale.
Benchmarks
State-of-the-Art Performance
Gemini 3.1 Pro sets new records on the most rigorous AI benchmarks, demonstrating genuine reasoning breakthroughs.
Reasoning
77.1%
ARC-AGI-2
More than double Gemini 3 Pro's score on novel logic pattern solving — a verified frontier reasoning breakthrough.
Knowledge
44.4%
Humanity's Last Exam
Record score on advanced domain-specific knowledge, surpassing GPT-5.2 (34.5%) and Gemini 3 Pro (37.5%).
Engineering
1.27x
RE-Bench (ML R&D)
Human-normalized score of 1.27 on ML research engineering tasks, cutting LLM fine-tuning runtime from 300s to 47s.
Comparison
Superior
vs Gemini 2.5 Pro
Significantly outperforms Gemini 2.5 Pro across benchmarks requiring enhanced reasoning and multimodal capabilities.
Safety
Improved
Multilingual Safety
Enhanced multilingual safety scores and refined refusal tone relative to Gemini 3.0 Pro.
Context
1M tokens
Context Utilization
Industry-leading 1M token context window for processing entire codebases, research corpora, and complex workflows.
Key Features
Gemini 3.1 Pro Key Features
A comprehensive feature set designed for developers, enterprises, and researchers tackling the most complex AI challenges.
1M Token Context Window
Process entire codebases, long research corpora, or deeply nested workflows — keeping large task graphs in memory across complex sessions.
64K Token Output
Generate extensive, detailed responses — from full system designs to comprehensive reports — without fragmentation.
Code-Based Animation
Generate website-ready animated SVGs directly from text prompts, producing crisp, scalable visuals with tiny file sizes.
Complex System Synthesis
Bridge complex APIs and user-friendly design — build live dashboards, configure telemetry streams, and visualize real-time data.
Interactive 3D Design
Code immersive 3D experiences with hand-tracking and generative audio — prototype sensory-rich interfaces with ease.
Safety & Reliability
Improved multilingual safety, refined refusal tone, and rigorous Frontier Safety Framework compliance for production deployments.
Why Use
Why use Gemini 3.1 Pro
Designed for complex enterprise tasks that require stronger reasoning, multimodal understanding, and long-context execution.
Native multimodal workflows
Gemini 3.1 Pro can reason across text, image, audio, video, and code in one pipeline for richer task execution.
1M context for large artifacts
Teams can analyze long reports, large code repositories, and multi-step plans with fewer context breaks.
How to Use
How to use Gemini 3.1 Pro in practice
Use a structured approach to improve output quality for API integration, research tasks, and enterprise automation.
Map tasks by modality
Define which parts need text, image, audio, or code reasoning before prompting to reduce unnecessary token usage.
Set stage-based prompts
Split requests into analysis, synthesis, and final output stages to keep complex workflows more controllable.
Run model-level comparisons
Compare Gemini and alternatives on your own benchmark tasks and track quality, latency, and cost by scenario.
Related models
Compare other frontier models on Fluxchat.
GPT-5.4
OpenAI's professional model for coding, reasoning, tools, and long-context work
Claude Sonnet 4.6
Anthropic's strongest Sonnet with near Opus-level intelligence
Claude Opus 4.6
Anthropic's most intelligent model with 1M context
GLM-5
Zhipu AI's flagship model with open SOTA coding & agent performance
FAQ
Frequently Asked Questions
Have more questions? Contact our support team.
Gemini 3.1 Pro
Experience Gemini 3.1 Pro on Fluxchat
Try Google's most advanced reasoning model with breakthrough ARC-AGI-2 performance and 1M token context on Fluxchat.