Gemini 3.1 Pro Preview Benchmark Analysis

Gemini 3.1 Pro Preview currently shows benchmark results led by GPQA Diamond (3 / 180, score 94.30), LiveCodeBench (3 / 120, score 91.70), LiveBench (3 / 115, score 79.93). This page also compares it with 2 competitor models and 3 predecessor or same-series models, including performance and pricing views when available. 3 source links are attached for reference.

Google 在模型评估中提供了多项基准测试结果。Gemini 3.1 Pro 在以下测试中得分高于 Gemini 3 Pro（括号内为 Gemini 3 Pro 对应分数，所有结果基于 2026 年 2 月数据，使用指定思考模式或无工具设置）：

Humanity's Last Exam（学术推理，全集，文本+多模态，无工具）：44.4%（37.5%） ARC-AGI-2（抽象推理谜题，ARC Prize Verified）：77.1%（31.1%） GPQA Diamond（科学知识，无工具）：94.3%（91.9%） Terminal-Bench 2.0（代理终端编码，Terminus-2 harness）：68.5%（56.9%） SWE-Bench Verified（代理编码，单次尝试）：80.6%（76.2%） LiveCodeBench Pro（竞争性编程，Elo 分数）：2887（2439） MMMLU（多语言问答）：92.6%（91.8%） MRCR v2（长上下文，128k 平均）：84.9%（77.0%）

在多模态理解（MMMU-Pro，无工具）上得分为 80.5%（Gemini 3 Pro 为 81.0%）。在代理工具使用和长上下文等类别中，多数指标也有相应变化。Google 同时提供了与其他模型（如 Claude Sonnet 4.6、Opus 4.6 及 GPT 系列）的比较数据。模型卡指出，Gemini 3.1 Pro 在前沿安全框架评估中未达到关键能力水平（CCL）警戒阈值，包括 CBRN、cyber 等风险领域。安全评估显示，与 Gemini 3 Pro 相比，部分指标有小幅变化。

Benchmark Results

Gemini 3.1 Pro Preview

Benchmark Results

General Knowledge

7 evaluations

Benchmark / mode

Score

Rank/total

GPQA Diamond

High

94.30

3 / 180

MMLU

High

92.60

3 / 65

LiveBench

High

79.93

3 / 115

ARC-AGI-2

High

77.10

7 / 59

HLE

High

44.40

39 / 163

HLE

HighTools

51.40

18 / 163

ARC-AGI-3

High

3 / 6

Coding and Software Engineer

4 evaluations

Benchmark / mode

Score

Rank/total

LiveCodeBench

HighTools

91.70

3 / 120

SWE-bench Verified

HighTools

80.60

10 / 109

SWE-Bench Pro - Public

HighTools

54.20

25 / 45

DeepSWE

HighTools

10 / 10

Multimodal Understanding

1 evaluations

Benchmark / mode

Score

Rank/total

MMMU

High

80.50

11 / 28

常识推理

1 evaluations

Benchmark / mode

Score

Rank/total

Simple Bench

Standard Mode

79.60

2 / 63

Agent Level Benchmark

2 evaluations

Benchmark / mode

Score

Rank/total

τ²-Bench - Telecom

HighTools

99.30

1 / 35

τ²-Bench

HighTools

90.80

2 / 40

Math and Reasoning

3 evaluations

Benchmark / mode

Score

Rank/total

FrontierMath

High

36.90

11 / 60

FrontierMath - Tier 4

Standard Mode

16.70

20 / 80

FrontierMath - Tier 4

High

16.70

20 / 80

AI Agent - Information Search

1 evaluations

Benchmark / mode

Score

Rank/total

BrowseComp

HighToolsInternet

85.90

3 / 47

AI Agent - Tool Usage

3 evaluations

Benchmark / mode

Score

Rank/total

MCP-Atlas

HighTools

78.20

7 / 24

TerminalBench 2.1

HighTools

70.70

11 / 17

Terminal Bench 2.0

HighTools

68.50

8 / 46

Claw-style Agent Evaluation

1 evaluations

Benchmark / mode

Score

Rank/total

Pinch Bench

Thinking EnabledTools

86.70

10 / 37

Compare with other models

Competitor Comparison

Benchmark scores for Gemini 3.1 Pro Preview compared against top models in its class

Gemini 3.1 Pro PreviewClaude Opus 4.6 GPT-5.3

Benchmark categories:

The chart shows each model’s highest score per benchmark within the current filter. Out-of-100 benchmarks use raw heights; out-of-range benchmarks are scaled within that benchmark while labels keep the original scores.

12 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.

Benchmark	Gemini 3.1 Pro PreviewCurrent	Claude Opus 4.6
ARC-AGI-2 综合评估	77.10Thinking Level · High	66.30Extended Thinking
GPQA Diamond 综合评估	94.30Thinking Level · High	91.31Extended Thinking
HLE 综合评估	51.40Thinking Level · High ｜ Tools	53.00Extended Thinking ｜ Tools
MMLU 综合评估	92.60Thinking Level · High	91.05Extended Thinking
LiveCodeBench 编程与软件工程	91.70Thinking Level · High ｜ Tools	76.00Extended Thinking
SWE-bench Verified 编程与软件工程	80.60Thinking Level · High ｜ Tools	80.84Extended Thinking ｜ Tools
MMMU 多模态理解	80.50Thinking Level · High	77.30Extended Thinking ｜ Tools
Simple Bench 常识推理	79.60Standard Mode	67.60Standard Mode
τ²-Bench Agent能力评测	90.80Thinking Level · High ｜ Tools	91.89Extended Thinking ｜ Tools
τ²-Bench - Telecom Agent能力评测	99.30Thinking Level · High ｜ Tools	99.25Extended Thinking ｜ Tools
FrontierMath 数学推理	36.90Thinking Level · High	40.70Thinking Level · High
FrontierMath - Tier 4 数学推理	16.70Thinking Level · High	22.90Thinking Level · High

3 additional benchmarks remain in the chart above.

Standard API Pricing: Gemini 3.1 Pro Preview vs. Peer Models

Shows standard text input and output pricing side by side for each model. If extended-context pricing exists, the chart keeps the base rate and explains the threshold below.

Source: DataLearnerAI. Standard text prices shown here use the default supplier. · USD / 1M tokens

When a context threshold exists, the charted base price only applies within these limits:

Gemini 3.1 Pro Preview: Base price applies to <= 200K

Claude Opus 4.6: Base price applies to <= 200K

Model	Supplier	Standard input	Standard output	Base price applies to
Gemini 3.1 Pro Preview	Google Deep Mind	$2 / 1M tokens	$12 / 1M tokens	<= 200K
Claude Opus 4.6	Anthropic	$5 / 1M tokens	$25 / 1M tokens	<= 200K

Version History

How each version of the Gemini 3.1 Pro Preview series stacks up on benchmark tests

Gemini 3.1 Pro PreviewGemini 3.0 Pro (Preview 11-2025)Gemini 2.5-Pro Gemini 2.5 Pro Experimental 03-25

Benchmark categories:

5 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.· Click a row to view its trend chart.

Benchmark	Gemini 3.1 Pro PreviewCurrent	Gemini 3.0 Pro (Preview 11-2025)	Gemini 2.5-Pro	Gemini 2.5 Pro Experimental 03-25
LiveBench 综合评估	79.93Thinking Level · High	73.39Thinking Level · High	58.33Thinking Level · High	--
FrontierMath - Tier 4 数学推理	16.70Standard Mode	18.80Thinking Enabled	2.10Standard Mode	--
BrowseComp AI Agent - 信息收集	85.90Thinking Level · High ｜ Tools	59.20Thinking Level · High ｜ Tools	7.80Thinking Enabled ｜ Tools	--
MCP-Atlas AI Agent - 工具使用	78.20Thinking Level · High ｜ Tools	70.30Standard Mode ｜ Tools	--	--
Pinch Bench OpenClaw智能体能力综合测评	86.70Thinking Enabled ｜ Tools	70.70Thinking Enabled ｜ Tools	--	71.90Thinking Enabled ｜ Tools

Single-Benchmark Version Trend

Viewing: LiveBench · 综合评估

Benchmark

NormalNormal + ToolsThinkingThinking + ToolsDeepDeep + Tools

X-axis shows model and release date, Y-axis shows score; solid lines connect the same mode across versions, while dotted guides align modes within the same generation.

Standard API Pricing Across the Gemini 3.1 Pro Preview Series

Shows standard text input and output pricing side by side for each model. If extended-context pricing exists, the chart keeps the base rate and explains the threshold below.

Source: DataLearnerAI. Standard text prices shown here use the default supplier. · USD / 1M tokens

When a context threshold exists, the charted base price only applies within these limits:

Gemini 3.1 Pro Preview: Base price applies to <= 200K

Model	Supplier	Standard input	Standard output	Base price applies to
Gemini 3.1 Pro Preview	Google Deep Mind	$2 / 1M tokens	$12 / 1M tokens	<= 200K

Sources

blog.googleblog.google arcprize.orgarcprize.org epoch.aiepoch.ai