Gemma 3 4B Instruct

GoogleGemma開源權重Gemma · 商用許可

描述

Gemma 3 4B is a 4-billion-parameter vision-language model from Google, handling text and image input and generating text output. It features a 128K context window, multilingual support, and open weights. Suitable for question answering, summarization, reasoning, and image understanding tasks.

發布日期

2025-03-12

參數規模

4.0B

上下文長度

131K

支援模態

image, text

能力雷達圖

general

coding

reasoning

science

agents

multimodal

排行榜排名

領域	#排名	分數	來源
程式碼能力榜	533	5.0	AA
通用能力榜	550	14.0	AA
多模態榜	70	36.0	LS
科學能力	543	14.0	AA

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Biology

GPQANYU + Cohere + Anthropic (2023)

30.8%自報

Code

HumanEvalOpenAI (2021)

71.3%自報

LiveCodeBench

12.6%自報

Factuality

FACTS Grounding

70.1%自報

SimpleQA

4.0%自報

Finance

MMLU-Pro

43.6%自報

General

IFEvalGoogle Research (2023)

90.2%自報

Natural2Code

70.3%自報

MBPP

0.63 / 100自報

Global-MMLU-Lite

54.5%自報

MMMU (val)

48.8%自報

BIG-Bench Extra Hard

11.0%自報

Image To Text

DocVQADocVQA (2020)

75.8%自報

VQAv2 (val)

62.4%自報

TextVQA

57.8%自報

Language

BIG-Bench Hard

72.2%自報

WMT24++

46.8%自報

ECLeKTic

4.6%自報

Math

GSM8k

89.2%自報

MATH

75.6%自報

MathVista-Mini

50.0%自報

HiddenMath

43.0%自報

Multimodal

AI2D

74.8%自報

ChartQAMasry et al. (2022)

68.8%自報

InfoVQA

50.0%自報

Reasoning

Bird-SQL (dev)

36.3%自報

AA 評測指數

(Artificial Analysis)

Math Index(Artificial Analysis)

12.7

Coding Index(Artificial Analysis)

2.7

Intelligence Index(Artificial Analysis)

1.0

Math 500(OpenAI (2024), subset of Hendrycks et al. MATH (2021))

0.8

Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))

0.4

Gpqa(NYU + Cohere + Anthropic (2023))

0.3

Ifbench(Google Research (2023))

0.3

Aime 25(MAA (Mathematical Association of America))

0.1

Livecodebench(UC Berkeley + MIT + Cornell (2024))

0.1

Scicode(UIUC + Argonne National Lab (2024))

0.1

Lcr(Artificial Analysis)

0.1

Aime(MAA (Mathematical Association of America))

0.1

Hle(Center for AI Safety + Scale AI (2025))

0.1

Tau2(Sierra + U Toronto + Vector Institute (2025))

0.0

Terminalbench Hard(Stanford × Laude Institute (2026))

0.0

Tau Banking

0.0

Terminalbench V2 1

0.0

LLM Stats 分類評分

(LLM Stats (zeroeval))

Structured Output

Instruction Following

Image To Text

Grounding

Math

Multimodal

Vision

Reasoning

General

Healthcare

Legal

Language

Factuality

Finance

Code

Physics

Biology

Chemistry

定價

輸入價格免費

輸出價格免費

混合價格(3:1)免費

速度

Tokens/秒0.0

首Token延遲0.00s

首回答延遲0.00s

供應商價格排行

2 個供應商

最便宜: OpenRouter最貴: Kilo Gateway

供應商輸入輸出

1OpenRouter最便宜

$0.05

$0.1

2Kilo Gateway

$0.05

$0.1

比較該模型在不同 API 供應商之間的定價。

外部連結

LLM Stats Artificial Analysis