Gemini 3.1 Pro

Name: Gemini 3.1 Pro
Author: Google

Multimodal

Google

Gemini 3.1 Pro is the latest model in the Gemini 3 series. Excels at complex tasks requiring extensive world knowledge and advanced multimodal reasoning. Gemini 3.1 Pro uses dynamic thinking by default to process prompts and has a 1 million token input context window with 64k output tokens.

Key Specifications

Parameters

Context

1.0M

Release Date

February 19, 2026

Average Score

63.3%

API Documentation Results Blog

Timeline

Key dates in the model's history

Announcement

February 19, 2026

Last Update

February 20, 2026

Today

October 5, 2026

Technical Specifications

Parameters

Training Tokens

Knowledge Cutoff

January 1, 2025

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$2.50

Output (per 1M tokens)

$15.00

Max Input Tokens

1.0M

Max Output Tokens

65.5K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-bench Verified

SWE-bench Verified — benchmark for evaluation abilities model solve real tasks from GitHub- • Self-reported

80.6%

Reasoning

Logical reasoning and analysis

GPQA

GPQA Diamond — benchmark for evaluation abilities model answer on questions level PhD by and • Self-reported

94.3%

Other Tests

Specialized benchmarks

ARC-AGI v2

ARC-AGI v2 — benchmark for evaluation abilities to reasoning and generalization • Self-reported

77.1%

MMMLU

MMMLU — version MMLU for evaluation knowledge model on languages • Self-reported

92.6%

CharXiv-R

CharXiv-R — benchmark for evaluation abilities model understand and reason about and • Self-reported

85.9%

MMMU-Pro

MMMU-Pro — version MMMU for evaluation on level experts • Self-reported

80.5%

HLE

HLE (Humanity's Last Exam) — benchmark from questions, experts for verification knowledge AI • Self-reported

51.4%

APEX-Agents

• Self-reported

33.5%

BrowseComp

Search + Python + Browse • Self-reported

85.9%

DeepSWE 1.1

DeepSWE v1.1 leaderboard, mini-swe-agent harness, high effort; Pass@1 12% ± 2%. • Verified

12.0%

Finance Agent v2

• Verified

43.0%

FrontierSWE

Gemini CLI • Verified

40.0%

Legal Agent Benchmark

All-pass rate (Harvey held-out set) • Verified

LiveBench

2026-01-08, High • Verified

79.9%

LiveCodeBench Pro

Elo Rating • Self-reported

96.2%

MCP Atlas

• Self-reported

69.2%

MRCR v2 (8-needle)

1M (pointwise) • Self-reported

26.3%

SciCode

• Self-reported

59.0%

SWE-Bench Pro

Single attempt • Self-reported

54.2%

t2-bench

Telecom • Self-reported

99.3%

Terminal-Bench 2.0

Terminus-2 harness • Self-reported

68.5%

License & Metadata

License

proprietary

Announcement Date

February 19, 2026

Last Updated

February 20, 2026

Articles about Gemini 3.1 Pro

Every Frontier AI Model Scored Under 1% on ARC-AGI-3. Humans Got 100%.

Chollet's new benchmark drops the same week Jensen Huang declared AGI. GPT-5.4 scored 0.26%. Claude Opus 4.6 scored 0.25%. The gap with humans is 99+ points.

March 27, 2026

9 min

Chollet to Jensen Huang: 'You Don't Have AGI. You Have Expensive Autocomplete.'

The ARC-AGI creator fires back at NVIDIA's AGI claims with a benchmark that every frontier model fails. His argument is simple: if humans can do it cold, AGI should too.

Google

Released:Jul 2026

Price:$1.50/1M tokens

Gemini 3.5 Flash-Lite

Google

Released:Jul 2026

Price:$0.30/1M tokens

Gemini 2.5 Flash-Lite

Google

Best score:0.6 (GPQA)

Released:Jun 2025

Price:$0.10/1M tokens

Gemini 1.5 Pro

Google

Best score:0.9 (MMLU)

Released:May 2024

Price:$2.50/1M tokens

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.