Robots Atlas>ROBOTS ATLAS
Gemini 3.1 Pro
AI Models›Gemini

Gemini 3.1 Pro

3.1 Pro · Family: Gemini
Gemini 3.1 Pro is Google DeepMind's most capable general-purpose model from the Gemini 3.1 family – a multimodal model with enhanced reasoning, agentic coding, and long-context capabilities.
⏳ Preview⏳ Limited accessLLMMultimodalReasoning modelTool-using model📁 Gemini
Context window
1M
tokens
Max output
65,536
tokens
Release date
29 April 2026
🔬Google DeepMindResearch lab🏢GoogleOwner
Access:APIHostedDeployment:☁ Cloud

Overview

Gemini 3.1 Pro is an AI model developed by Google DeepMind, announced on April 29, 2026 as the successor to Gemini 3 Pro. It is a multimodal model supporting text, image, video, audio, and PDF document inputs, generating text as output.

The model has a context window of up to 1 million tokens and a maximum output of 64,000 tokens. It supports tools including function calling, structured output, search as a tool, and code execution. Available via Gemini App, Google Cloud/Vertex AI, Google AI Studio, Gemini API, Google AI Mode, and Google Antigravity.

Knowledge cutoff: January 2025. Available in preview. Particularly strong in agentic coding (SWE-Bench 80.6%), reasoning (ARC-AGI-2 77.1%), and long-context tasks (MRCR v2 128k: 84.9%).

Classification
LLMMultimodalReasoning modelTool-using model
Family: Gemini
Applications
CodingContent generationDocument analysisWorkflow automationWriting assistanceQ&A / Question answeringResearch assistanceData analysis
Access & deployment
APIHosted
Cloud
Weights: Closed
Key parameters
📏 Context: 1M
✓ Tools
📥 Input: text, image, audio, video…
Platforms
Vertex AI

Technical specification

Context window
1M
tokens
Max output tokens
65,536
tokens per response
Knowledge cutoff
1 Jan 2025
Knowledge boundary
License
proprietary
Hardware requirements
Available only through Google cloud infrastructure (Gemini API, Vertex AI, Google AI Studio, Google Antigravity).
Features:✓ Tool use
Modalities
⬇ Input
textimageaudiovideodocuments
⬆ Output
textcode

Capabilities and applications

Native model capabilities
Reasoning
Category: reasoning
Multi-step reasoning
Category: reasoning
Long context
Category: reasoning
Multimodal understanding
Category: multimodal
Coding
Category: coding
Function Calling
Category: planning
Structured output
Category: structured_generation
Audio understanding
Category: audio
Image understanding
Category: vision
Video Understanding
Category: video
Chart understanding
Category: vision
Diagram reasoning
Category: reasoning
OCR
Category: vision
Multilingual
Category: language
Planning
Category: planning
Streaming output
Category: reasoning
Interleaved Multimodal Input
Category: reasoning
Application domains
CodingContent generationDocument analysisWorkflow automationWriting assistanceQ&A / Question answeringResearch assistanceData analysisSearch assistance

Benchmark results

18 benchmarks
Humanity's Last Exam
accuracy · No tools, Gemini 3.1 Pro Thinking (High)
44.4%%
📄 https://deepmind.google/models/gemini/pro/
Full set (text + MM). No tools.
Humanity's Last Exam
accuracy · Search (blocklist) + Code, Gemini 3.1 Pro Thinking (High)
51.4%%
📄 https://deepmind.google/models/gemini/pro/
Full set (text + MM). With search and code execution.
ARC-AGI-2
accuracy · ARC Prize Verified, Gemini 3.1 Pro Thinking (High)
77.1%%
📄 https://deepmind.google/models/gemini/pro/
Abstract reasoning puzzles, ARC Prize verified.
GPQA Diamond
accuracy · No tools, Gemini 3.1 Pro Thinking (High)
94.3%%
📄 https://deepmind.google/models/gemini/pro/
Scientific knowledge, no tools.
Terminal-Bench 2.0
accuracy · Terminus-2 harness, Gemini 3.1 Pro Thinking (High)
68.5%%
📄 https://deepmind.google/models/gemini/pro/
Agentic terminal coding.
SWE-Bench Verified
accuracy · Single attempt, Gemini 3.1 Pro Thinking (High)
80.6%%
📄 https://deepmind.google/models/gemini/pro/
Agentic coding, single attempt.
SWE-Bench Pro (Public)
accuracy · Single attempt, Gemini 3.1 Pro Thinking (High)
54.2%%
📄 https://deepmind.google/models/gemini/pro/
Diverse agentic coding tasks.
LiveCodeBench Pro
elo · Gemini 3.1 Pro Thinking (High)
2887 EloElo
📄 https://deepmind.google/models/gemini/pro/
Competitive coding problems from Codeforces, ICPC, and IOI.
SciCode
accuracy · Gemini 3.1 Pro Thinking (High)
59%%
📄 https://deepmind.google/models/gemini/pro/
Scientific research coding.
APEX-Agents
accuracy · Gemini 3.1 Pro Thinking (High)
33.5%%
📄 https://deepmind.google/models/gemini/pro/
Long horizon professional tasks.
τ2-bench (Retail)
accuracy · Retail, Gemini 3.1 Pro Thinking (High)
90.8%%
📄 https://deepmind.google/models/gemini/pro/
Agentic tool use – retail.
τ2-bench (Telecom)
accuracy · Telecom, Gemini 3.1 Pro Thinking (High)
99.3%%
📄 https://deepmind.google/models/gemini/pro/
Agentic tool use – telecom.
MCP Atlas
accuracy · Gemini 3.1 Pro Thinking (High)
69.2%%
📄 https://deepmind.google/models/gemini/pro/
Multi-step workflows using MCP.
BrowseComp
accuracy · Search + Python + Browse, Gemini 3.1 Pro Thinking (High)
85.9%%
📄 https://deepmind.google/models/gemini/pro/
Agentic search.
MMMU-Pro
accuracy · No tools, Gemini 3.1 Pro Thinking (High)
80.5%%
📄 https://deepmind.google/models/gemini/pro/
Multimodal understanding and reasoning.
MMMLU
accuracy · Gemini 3.1 Pro Thinking (High)
92.6%%
📄 https://deepmind.google/models/gemini/pro/
Multilingual Q&A.
MRCR v2 (8-needle, 128k)
accuracy · 128k average, Gemini 3.1 Pro Thinking (High)
84.9%%
📄 https://deepmind.google/models/gemini/pro/
Long context performance.
MRCR v2 (8-needle, 1M)
accuracy · 1M pointwise, Gemini 3.1 Pro Thinking (High)
26.3%%
📄 https://deepmind.google/models/gemini/pro/
Very long context performance (1M tokens).

Pricing

Technical architecture

Deployment and security

☁ Available on platforms
☁Vertex AIPlatform
🔒 Security / Enterprise
✓ Verified enterprise information

Gemini 3.1 Pro dostępny w Vertex AI i Gemini Enterprise. Model card dostępny publicznie pod adresem deepmind.google/models/model-cards/gemini-3-1-pro.

Updated: 1 May 2026↗ Security documentation

Sources and related pages