NOUMENOpen app
NOUMEN

Google · Open weights

Gemma 3 4B IT

Added Jan 1, 2024Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Gemma 3 has a large, 128K context window, multilingual support in over 140 languages, and is available in more sizes than previous versions.

  • PDF input
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.20
Output
$0.20

Reference prices as published by the model provider. API example

Specifications

Context window
128K
Max output
8.2K
Average output (7 days)
115 tokens
Longer than 11% of 468 models

Side by side

Gemma 3 4B IT compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Gemma 3 4B ITGoogleThis modelGemma 4 31B ThinkingGoogleRemove
ProviderGoogleGoogle
Discovery tierOpen weightsOpen weights
AddedJan 1, 2024Apr 2, 2026
Limits
Context window128K262.14KBest
Max output8.19K131.07KBest
Average output115 tokens1.1K tokens
Pricing, USD per 1M tokens
Input$0.20$0.10Best
Output$0.20Best$0.34
Cache read—$0.10
Performance
Throughput—27.7 tok/s
Time to first token—7.2 s
Request success (24h)100.0%Best99.7%
Benchmark indexes
Intelligence Index4.814.7Best
Coding Index2.743.4Best
Math Index12.7—
Capabilities
ReasoningNoYes
ToolsNoYes
VisionNoYes
Structured outputNoYes
PDF inputYesNo
Video inputNoYes
Reasoning effort——

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Gemma 3 4B IT
  • Gemma 4 31B Thinking

Intelligence Index

  • Gemma 3 4B IT4.8
  • Gemma 4 31B Thinking14.7

Gemma 3 4B IT: Better than 0% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Gemma 3 4B IT: 2.7
    • Gemma 4 31B Thinking: 43.4Top

  • Math Index

    Composite of the math evaluations

    • Gemma 3 4B IT: 12.7
    • Gemma 4 31B Thinking: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 4.9%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 0.5%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 368 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 619 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 6.0%

  • AA-LCRv1.1

    Reasoning across long documents

    • Gemma 3 4B IT: 6.7%
    • Gemma 4 31B Thinking: 69.7%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Gemma 3 4B IT: 5.3%
    • Gemma 4 31B Thinking: 23.6%Top

  • GPQA Diamond

    Graduate-level science questions

    • Gemma 3 4B IT: 29.1%
    • Gemma 4 31B Thinking: 85.7%Top

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Gemma 3 4B IT: 41.7%
    • Gemma 4 31B Thinking: —

  • IFBench

    Following precise output instructions

    • Gemma 3 4B IT: 28.3%
    • Gemma 4 31B Thinking: 75.6%Top

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 0.0%

  • SciCode

    Scientific computing in Python

    • Gemma 3 4B IT: —
    • Gemma 4 31B Thinking: 45.5%

  • LiveCodeBench

    Recent competitive programming problems

    • Gemma 3 4B IT: 11.2%
    • Gemma 4 31B Thinking: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Gemma 3 4B IT: 12.7%
    • Gemma 4 31B Thinking: —

  • AIME

    Competition mathematics

    • Gemma 3 4B IT: 6.3%
    • Gemma 4 31B Thinking: —

  • MATH-500

    Competition math problems

    • Gemma 3 4B IT: 76.6%
    • Gemma 4 31B Thinking: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Gemma 3 4B IT: 0.8%
    • Gemma 4 31B Thinking: 36.4%Top

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Gemma 3 4B IT: 5.0%
    • Gemma 4 31B Thinking: 59.9%Top

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Gemma 3 4B IT: 6.7%
    • Gemma 4 31B Thinking: 69.7%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Request success (24h)
100.0%
Reliable
Hourly request success · last 72 hoursLowest hour 66.6%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "unsloth/gemma-3-4b-it",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="unsloth/gemma-3-4b-it",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Google

Other Google models, starting with the same family.