Saltar al contenido
MModelosIA
CatálogoLocalesBenchmarksHerramientasGuíaBlog

Open-weight

Modelos locales

Fichas open-weight del catálogo. La VRAM es una estimación para cuantización Q4; confirma el runtime (Ollama, llama.cpp, vLLM) y la variante exacta antes de comprar hardware.

ModeloEmpresaContextoVRAM est.Licencia
DeepSeek V4 ProDeepSeek1.048576M~960 GB (Q4)MITDetalle
Kimi K3Moonshot AI1.048576M~1680 GB (Q4)Open-weightDetalle
Meta Muse Spark 1.2Meta500KVariableLlama CommunityDetalle
Qwen 3.8 27BAlibaba (Qwen)128K~16 GB (Q4)Apache-2.0Detalle
GLM-5.2Zhipu AI1MVariableOpen-weightDetalle
GLM-5Zhipu AI205KVariableOpen-weightDetalle
MiMo-V2.5-ProXiaomi256KVariableApache-2.0Detalle
Meta Muse Glimmer 30BMeta256K~18 GB (Q4)Llama CommunityDetalle
Qwen3.5 (open-weight)Alibaba (Qwen)131KVariableApache-2.0Detalle
DeepSeek V3DeepSeek128K~411 GB (Q4)MITDetalle
Llama 4 MaverickMeta1.048576M~240 GB (Q4)Llama 4 CommunityDetalle
Mistral Large 3Mistral AI256K~74 GB (Q4)Apache-2.0Detalle
Phi-4 (familia)Microsoft16K~8 GB (Q4)MITDetalle
MiMo-V2.6-ProXiaomi—~612 GB (Q4)Open-weightDetalle
GLM-5.3 (max)Zhipu AI—~452 GB (Q4)Open-weightDetalle
GLM 5.3 FlashZhipu AI—~192 GB (Q4)Open-weightDetalle
DeepSeek V4.1 Flash (Reasoning, Max Effort)DeepSeek—~331 GB (Q4)Open-weightDetalle
K2 Horizon 375B A23BInstitute of Foundation Models—~225 GB (Q4)Open-weightDetalle
Inkling (xhigh)Thinking Machines1.048576M~585 GB (Q4)Open-weightDetalle
Muse Glimmer (high)Meta—~18 GB (Q4)Open-weightDetalle
Mistral SmallMistral AI128K~14 GB (Q4)Apache-2.0Detalle
Hunyuan LargeTencent256KVariableTencent HunyuanDetalle
MiniMax-M3MiniMax1.048576M~257 GB (Q4)Open-weightDetalle
Nemotron 3 Ultra 550B A55B (Reasoning)NVIDIA—~330 GB (Q4)Open-weightDetalle
Mistral Medium 3.5Mistral AI262K~77 GB (Q4)Open-weightDetalle
ModelosIA

Panel de modelos de IA para programadores. Precios, contexto y benchmarks en un solo sitio.

Los precios y ventanas de contexto cambian a menudo. Verifica siempre en la documentación oficial del proveedor antes de presupuestar. Datos actualizados a septiembre de 2026.

Navegar
  • Catálogo
  • Locales
  • Benchmarks
  • Herramientas
  • Guía
  • Blog
Proyecto
  • Jesús P. Rodríguez
  • GitHub
  • LinkedIn
  • modelosia.org
v3.0.0 · catálogo estático, sin backendHecho para elegir más rápido, no para vender un único modelo.

Web creada por webdespega