초소형 오픈 웨이트 AI 모델 비교(≤4B)

파라미터 수가 4B개 이하인 오픈 웨이트 AI 모델입니다. 일반적으로 필요한 리소스가 가장 적은 모델입니다.

웨이트를 다운로드할 수 있는 모델을 오픈 웨이트 모델(흔히 오픈 소스라고도 함)로 간주합니다. 자체 인프라에서 호스팅할 수 있으며 미세 조정 등을 통해 모델을 맞춤 설정할 수 있습니다.

방법론을 비롯한 자세한 내용은 FAQ에서 확인하세요.

AI9Stars 로고G9v3-3B과 Alibaba 로고Qwen3 4B 2507은 파라미터 수가 ≤4B개인 초소형 오픈 웨이트 모델 중 지능이 가장 높으며 OpenBMB 로고MiniCPM5-1B과 OpenBMB 로고MiniCPM5-1B이 뒤를 잇습니다.

주요 내용

개방성

Artificial Analysis Openness Index · Higher is better

지능

Artificial Analysis Intelligence Index · Higher is better

총 파라미터 수

학습 가능한 파라미터 수(단위: 십억 개)

개방성

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open)
Reasoning models are indicated by a lightbulb icon

지능

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR
Estimate (independent evaluation forthcoming)
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better
GDPval-AA v2

Agentic real-world work tasks, (Elo-500)/2000

𝜏³-Banking

Agentic tool use

Terminal-Bench v2.1

Agentic coding & terminal use

SciCode

Coding

Humanity's Last Exam

Reasoning & knowledge

GPQA Diamond

Scientific reasoning

CritPt

Physics reasoning

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

AA-LCR

Long context reasoning

AA-Briefcase

Agentic knowledge work, Elo

No data available
AutomationBench-AA

Agentic SaaS workflows

No data available
Harvey LAB-AA

Legal agentic work, criterion pass rate

No data available
EnterpriseOps-Gym-AANew

Agentic business operations

No data available
IFBench

Instruction following

APEX-Agents-AA

Long-horizon agentic tasks

No data available
ITBench-AA

Kubernetes incident root-cause analysis

No data available
MMMU-Pro

Visual reasoning

Reasoning models are indicated by a lightbulb icon

While model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

크기

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference
Reasoning models are indicated by a lightbulb icon

The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.

The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.

Intelligence Index vs. Active Parameters

Artificial Analysis Intelligence Index · Active parameters at inference time
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.

Intelligence Index vs. Total Parameters

Artificial Analysis Intelligence Index · Size in parameters (billions)
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.

컨텍스트 창

Context Window

Context window: tokens limit · Higher is better
Reasoning models are indicated by a lightbulb icon

Larger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

상세 정보

웨이트
제공업체 벤치마크
G9v3-3B
AI9Stars 로고AI9Stars
16
3B
131k
-
-
🤗
AI9Stars
보기
MiniCPM5-1B (Reasoning)
OpenBMB 로고OpenBMB
12
1B
128k
-
-
🤗
-
보기
MiniCPM5-1B (Non-reasoning)
OpenBMB 로고OpenBMB
12
1B
128k
-
-
🤗
-
보기
Nanbeige4.1-3B
Nanbeige 로고Nanbeige
11
3.9B
256k
-
-
🤗
-
보기
NVIDIA Nemotron 3 Nano 4B
NVIDIA 로고NVIDIA
9
4.0B
262k
-
-
🤗
-
보기
Qwen3.5 2B (Reasoning)
Alibaba 로고Alibaba
8
2.3B
262k
-
-
🤗
-
보기
Ministral 3 3B
Mistral 로고Mistral
7
3B
256k
$0.1
206
🤗
MistralAmazon Bedrock
보기
Phi-4 Mini Instruct
Microsoft 로고Microsoft
6
3.8B
128k
-
43
🤗
Microsoft Azure
보기
Qwen3.5 2B (Non-reasoning)
Alibaba 로고Alibaba
6
2.3B
262k
-
-
🤗
-
보기
Qwen3.5 0.8B (Reasoning)
Alibaba 로고Alibaba
5
0.9B
262k
-
-
🤗
-
보기
Granite 4.1 3B
IBM 로고IBM
5
3B
131k
-
-
🤗
-
보기
MiniCPM-V 4.6 1.3B
OpenBMB 로고OpenBMB
4
1.3B
262k
-
-
🤗
-
보기