微型开放权重 AI 模型比较（≤4B）

Name: 开放性
Creator: Artificial Analysis
License: https://artificialanalysis.ai/docs/legal/Terms-of-Use.pdf

参数量不超过 4B 的开放权重 AI 模型。就资源需求而言，它们通常是最小的模型。

如果模型权重可供下载，我们便将其视为开放权重模型。这样用户就可以在自己的基础设施上自行托管，并通过微调等方式定制模型。

如需了解包括方法论在内的更多详情，请参阅常见问题。

K2 Horizon 3.7B 和

MiniCPM5-2B 是智能得分最高的微型开放权重模型，定义为参数量不超过 4B 的模型，其次是

G9v3-3B 和

Granite 4.2 3B。

开放性

Artificial Analysis Openness Index · Higher is better

智能

Artificial Analysis Intelligence Index · Higher is better

开放性

Artificial Analysis Openness Index：得分

Openness Index 以 0 至 100 的标准化评分衡量模型的开放程度（分数越高，越开放）

智能

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.3.2 incorporates 10 evaluations: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1

Estimate (independent evaluation forthcoming)

Artificial Analysis Intelligence Index v4.3.2 includes: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Evaluations

Evaluations measured independently by Artificial Analysis · AA-PersonalityBench scores are percentile ranks from 1 to 100

AA-Briefcase v1.1

Agentic knowledge work, Elo

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

AutomationBench-AA

Agentic SaaS workflows

GDPval-AA v2.1

Agentic real-world work tasks, Elo

Humanity's Last Exam

Reasoning & knowledge

GDP.pdf

Professional document reasoning, All-pass

Terminal-Bench 4.0

Agentic coding & terminal use

AA-LCR v1.1

Long context reasoning

SciCodeUnder review

Coding

CritPt

Physics reasoning

Terminal-Bench-Science 0.1New

Agentic scientific research workflows in a terminal

No data available

FrontierCode 1.1New

Agentic software engineering in open-source repositories

No data available

AA-AnalystAgent

Quantitative analysis on spreadsheets & documents

No data available

Harvey LAB-AA v1.1Updated

Legal agentic work, Hallucination-Gated All-Pass Rate

No data available

EnterpriseOps-Gym-AA

Agentic business operations

No data available

AA-3DNew

Agentic interactive 3D experiences, Elo

No data available

ITBench-AA

Kubernetes incident root-cause analysis

No data available

MLCR-AAUpdated

Medical long context reasoning

No data available

MMMU-Pro

Visual reasoning

AA-PersonalityBench: Expressiveness ScoreNew

Rich, varied, well-crafted language

No data available

AA-PersonalityBench: Organization ScoreNew

Well-structured, scannable, answer-first

No data available

AA-PersonalityBench: Transparency ScoreNew

Truthful, acknowledges uncertainty, non-sycophantic

No data available

AA-PersonalityBench: Slop ScoreNew

Use of filler and generic language

No data available

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

规模

模型规模：总参数量与活跃参数量

Comparison between total model parameters and parameters active during inference (billions)

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 与活跃参数

Artificial Analysis Intelligence Index · Active parameters at inference time (billions)

Most attractive quadrant

Pareto line

Intelligence Index 与总参数量

Artificial Analysis Intelligence Index · Size in parameters (billions)

Most attractive quadrant

Pareto line

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.

上下文窗口

Context window: tokens limit · Higher is better

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).

							权重		服务商基准测试
K2 Horizon 3.7B	Institute of Foundation Models	16	3.7B	524k	-	-	🤗	-	查看
MiniCPM5-2B	OpenBMB	12	2.6B	131k	-	-	🤗	-	查看
G9v3-3B	AI9Stars	11	3B	131k	-	-	🤗	-	查看
Granite 4.2 3B	IBM	9	3B	131k	$0.0	99	🤗		查看
MiniCPM5-1B (Reasoning)	OpenBMB	9	1B	128k	-	-	🤗	-	查看
MiniCPM5-1B (Non-reasoning)	OpenBMB	9	1B	128k	-	-	🤗	-	查看
Nanbeige4.1-3B	Nanbeige	8	3.9B	256k	-	-	🤗	-	查看
LFM2.5-2.6B	Liquid AI	8	2.7B	128k	-	-	🤗	-	查看
Qwen3.5 2B (Reasoning)	Alibaba	7	2.3B	262k	-	-	🤗	-	查看
Phi-4 Mini Instruct	Microsoft	6	3.8B	128k	$0.1	46	🤗		查看
Qwen3.5 2B (Non-reasoning)	Alibaba	6	2.3B	262k	-	-	🤗	-	查看
Qwen3.5 0.8B (Reasoning)	Alibaba	6	0.9B	262k	-	-	🤗	-	查看

微型开放权重 AI 模型比较（≤4B）

开放性

智能

总参数量

开放性

Artificial Analysis Openness Index：得分

智能

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index

Evaluations

Intelligence Evaluation Relevance

Artificial Analysis Intelligence Index

规模

模型规模：总参数量与活跃参数量

Total Parameters

Active Parameters at Inference Time

Intelligence Index 与活跃参数

Artificial Analysis Intelligence Index

Active Parameters at Inference Time

Intelligence Index 与总参数量

Artificial Analysis Intelligence Index

Total Parameters

上下文窗口

上下文窗口

Context Window for RAG

Context Window

更多详情