Home/Models/Gemini 3.1 Pro Preview
Google DeepMind logo

Gemini 3.1 Pro Preview

by Google DeepMind · Released Feb 2026

Multimodal1M ContextPreview
90.0
avg score
Rank #7
Compare
Top 3% of all models
Context
1.0M tokens (~524 books)
Input $/1M
$2.00
Output $/1M
$12.00
Type
multimodal
License
Proprietary
Benchmarks
23 tested
Data updated today
About

Gemini 3.1 Pro Preview is Google’s frontier reasoning model, delivering enhanced software engineering performance, improved agentic reliability, and more efficient token usage across complex workflows. Building on the multimodal foundation...

Tested on 23 benchmarks with 60.6% average. Top scores: Chatbot Arena Elo — Overall (1492.6%), Chatbot Arena Elo — Coding (1455.7%), ARC-AGI (98.0%).

Looking for similar performance at lower cost?
Step 3.5 Flash scores 89.5 (99% as good) at $0.10/1M input · 95% cheaper
Capabilities
coding
75.4
#10 globally
reasoning
83.5
#5 globally
math
49.7
#77 globally
knowledge
53.9
#78 globally
agentic
33.5
#13 globally
speed
95.4
#2 globally
Benchmark Scores
Compare All
Tested on 23 benchmarks · Ranked across 7 categories
Score Distribution (all 233 models)
0255075100
▲ You are here
Terminal Bench

Complex terminal-based engineering tasks. Models must use command-line tools, navigate filesystems, and debug systems through shell interaction.

78.4
SWE-Bench verified

Real-world software engineering tasks from GitHub issues. Models must diagnose bugs and write patches that pass test suites. Human-verified subset of SWE-bench.

75.6
WeirdML

Unusual and adversarial machine learning challenges. Tests robustness of reasoning about edge cases in ML systems.

72.1
ARC-AGI

Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.

98.0
ARC-AGI-2

ARC-AGI 2, harder sequel to ARC. More complex abstract reasoning patterns that test generalization ability beyond training data.

77.1
SimpleBench

Deceptively simple questions that humans find easy but AI models often get wrong. Tests common sense and reasoning gaps.

75.5
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

95.6
FrontierMath-2025-02-28-Private

Original research-level math problems created by professional mathematicians. Problems are unpublished and cannot be memorized.

36.9
FrontierMath-Tier-4-2025-07-01-Private

Hardest tier of FrontierMath. Problems at the frontier of human mathematical ability, many unsolved by most mathematicians.

16.7
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Links
Documentation
BenchGecko API
gemini-3-1-pro-preview
Specifications
  • Typemultimodal
  • Context1.0M tokens (~524 books)
  • ReleasedFeb 2026
  • LicenseProprietary
  • Statuspreview
  • Cost / Message~$0.016
Available On
Google DeepMind logoGoogle DeepMind$2.00
Share & Export
Tweet
Gemini 3.1 Pro Preview is a proprietary multimodal AI model by Google DeepMind, released in February 2026. It has an average benchmark score of 90.0. Context window: 1M tokens.