How much does Gemma 4 31B cost?

Gemma 4 31B costs $0.13 per million input tokens and $0.38 per million output tokens. For a typical conversation (~2,000 tokens), that's approximately $0.001 per message.

What benchmarks has Gemma 4 31B been tested on?

Gemma 4 31B has been evaluated on 8 benchmarks. Top scores: LiveBench — Mathematics: 73.9, LiveBench — Language: 71.3, LiveBench — If: 67.6.

Is Gemma 4 31B open source?

Yes, Gemma 4 31B is open source.

How does Gemma 4 31B compare to phi-3-mini 3.8B?

Gemma 4 31B has an average score of 68.2 while phi-3-mini 3.8B scores 68.3. phi-3-mini 3.8B slightly outperforms Gemma 4 31B overall. See full comparison →

Home/Models/Gemma 4 31B

Gemma 4 31B

Name: Gemma 4 31B
Price: 0.13 USD
Author: Google DeepMind

by Google DeepMind · Released Apr 2026

Open SourceMultimodal

68.2

avg score

Rank #42

Compare

Better than 82% of all models

Context

262K tokens (~131 books)

Input $/1M

$0.13

Output $/1M

$0.38

Type

multimodal

License

Open Source

Benchmarks

8 tested

Data updated today

About

Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function...

Tested on 8 benchmarks with 61.6% average. Top scores: LiveBench — Mathematics (73.9%), LiveBench — Language (71.3%), LiveBench — If (67.6%).

Capabilities

coding

50.2

#64 globally

reasoning

59.1

#32 globally

math

73.9

#26 globally

knowledge

61.6

#37 globally

language

69.5

#70 globally

Benchmark Scores

Compare All

Tested on 8 benchmarks · Ranked across 5 categories

Score Distribution (all 231 models)

0255075100

▲ You are here

codingCompare coding →

LiveBench — Coding

Regularly refreshed coding problems that avoid data contamination. New problems added monthly to prevent memorization.

60.3—

LiveBench — Agentic Coding

LiveBench coding tasks that require multi-step reasoning and tool use. Tests planning and execution of complex coding workflows.

40.0—

reasoningCompare reasoning →

LiveBench — Reasoning

Regularly refreshed reasoning problems testing logical deduction, spatial reasoning, and analytical thinking.

59.4—

LiveBench — Data Analysis

Fresh data analysis tasks testing ability to interpret tables, charts, and statistical data.

58.8—

mathCompare math →

LiveBench — Mathematics

Regularly updated math problems that test numerical reasoning, algebra, calculus, and combinatorics.

73.9—

Quick compare:

vs phi-3-mini 3.8B

vs Grok 3 Beta

vs GPT-4

Excellent (85+) Good (70-85) Average (50-70) Below (<50)

Similar Models

Links

Info

Research

Documentation

Community

Source Code

BenchGecko API

gemma-4-31b-it

Specifications

Typemultimodal
Context262K tokens (~131 books)
ReleasedApr 2026
LicenseOpen Source
StatusActive
Cost / Message~$0.001

Available On

Google DeepMind$0.13

Frequently Asked Questions

Gemma 4 31B is an open-source multimodal AI model by Google DeepMind, released in April 2026. It has an average benchmark score of 68.2. Context window: 262K tokens.

Benchmarks

LiveBench — Mathematics LiveBench — Language LiveBench — If LiveBench — Overall LiveBench — Coding

Google DeepMind · Provider All Models Compare Models Context Window · Glossary

Gemma 4 31B

Frequently Asked Questions

Related Models

Benchmarks

Related Pages