Comparación de modelos con pesos abiertos

Comparación y análisis de modelos de IA con pesos abiertos en métricas clave como calidad, rendimiento, velocidad de inferencia, ventana de contexto, número de parámetros y detalles de licencia.

Consideramos que un modelo tiene pesos abiertos (también denominado comúnmente «open source») cuando sus pesos están disponibles para descargar. Esto permite alojarlo en infraestructura propia y personalizarlo, por ejemplo, mediante ajuste fino.

Para más detalles sobre nuestra metodología, consulta nuestras FAQs.

Logo de KimiKimi K3 (max) y Logo de Z AIGLM-5.3 (max) son los modelos con pesos abiertos de mayor inteligencia, seguidos por Logo de AlibabaQwen3.8 2.4T A95B y Logo de Z AIGLM-5.3-Flash.

Aspectos destacados

Openness

Artificial Analysis Openness Index · Higher is better
Updated

Inteligencia

Artificial Analysis Intelligence Index · Higher is better

Parámetros totales

Parámetros entrenables en miles de millones

Openness

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open)

Progreso de los modelos con pesos abiertos

Progress in Open Weights vs. Proprietary Intelligence

Artificial Analysis Intelligence Index v4.2 incorporates 10 evaluations: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

Indicates whether the model weights are available. Models are labelled as 'Commercial Use Restricted' if commercial use is limited by conditions, and as 'Non-commercial' if the license prohibits commercial use.

Evolución de la inteligencia de los modelos de lenguaje con pesos abiertos por laboratorio

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

Evolución de la inteligencia de los modelos con pesos abiertos por tamaño

Artificial Analysis Intelligence Index v4.2 incorporates 10 evaluations: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

  • Tiny: Less than or equal to 4B parameters. These are usually the smallest models in terms of resource demand.
  • Small: Less than 40B parameters.
  • Medium: Between 40B-150B parameters.
  • Large: Over 150B parameters.

Inteligencia

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.2 incorporates 10 evaluations: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better
See more
AA-Briefcase

Agentic knowledge work, (Elo-500)/2000

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

GDP.pdfNew

Professional document reasoning, All-pass

CritPt

Physics reasoning

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

AA-LCR v1.1

Long context reasoning

AutomationBench-AA

Agentic SaaS workflows

Harvey LAB-AA

Legal agentic work, criterion pass rate

EnterpriseOps-Gym-AA

Agentic business operations

GPQA Diamond

Scientific reasoning

AA-AnalystAgent

Quantitative analysis on spreadsheets & documents

IFBench

Instruction following

APEX-Agents-AA

Long-horizon agentic tasks

ITBench-AA

Kubernetes incident root-cause analysis

MMMU-Pro

Visual reasoning

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

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

Tamaño

Intelligence Index por tamaño de modelo

Artificial Analysis Intelligence Index v4.2 incorporates 10 evaluations: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference

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

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

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

Artificial Analysis Intelligence Index v4.2 includes: AA-Briefcase, GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, 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.

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.

Ventana de contexto

Context Window

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).

Más detalles

Pesos
Benchmarks de proveedores
Kimi K3 (max)
Logo de KimiKimi
50
2.8T
104B activos en inferencia
1M
$2.3
41
🤗
FireworksFireworksModal
+13
Ver
GLM-5.3 (max)
Logo de Z AIZ AI
49
753B
40B activos en inferencia
1M
$0.9
83
🤗
ZaiSelf-hostedDeepInfra
+12
Ver
Qwen3.8 2.4T A95B
Logo de AlibabaAlibaba
47
2.4T
95B activos en inferencia
984k
$1.2
40
🤗
FireworksTogether AIDeepInfra
+3
Ver
GLM-5.3-Flash
Logo de Z AIZ AI
46
320B
18B activos en inferencia
1M
$0.1
46
🤗
ZaiBasetenNovita
+15
Ver
DeepSeek V4 Pro 0813 (Reasoning, Max Effort)
Logo de DeepSeekDeepSeek
42
1.6T
49B activos en inferencia
1M
$0.7
67
🤗
DeepSeekGMISiliconFlow
+6
Ver
Qwen3.8 27B (xhigh)
Logo de AlibabaAlibaba
41
27B
256k
$0.4
44
🤗
DeepInfraSelf-hostedCoreWeave
+4
Ver
K2 Horizon 375B A23B
Logo de MBZUAI Institute of Foundation ModelsMBZUAI Institute of Foundation Models
38
375B
23B activos en inferencia
524k
-
-
🤗
-
Ver
MiniMax-M3
Logo de MiniMaxMiniMax
36
428B
23B activos en inferencia
1M
$0.2
85
🤗
ParasailCoreWeaveTogether AI
+12
Ver
Inkling (xhigh)
Logo de Thinking MachinesThinking Machines
32
975B
41B activos en inferencia
1M
$0.7
71
🤗
Self-hostedDeepInfraBaseten
+3
Ver
Muse Glimmer (high)
Logo de MetaMeta
24
30B
131k
$0.2
100
🤗
Together AIDeepInfraFireworks
Ver
gpt-oss-120b (high)
Logo de OpenAIOpenAI
16
117B
5.1B activos en inferencia
131k
$0.2
176
🤗
GroqCloudflareScaleway
+17
Ver