开源模型选型指南

313 个开源模型，覆盖主流厂商。按参数规模、架构类型、VRAM 需求筛选，找到最适合你硬件条件的开源模型。

Large(>150B)

Medium(40-150B)

Small(4-40B)

187

Tiny(<=4B)

大小:

架构:

硬件:

共 313 个开源模型

排序:

VRAM 兼容性检查器

选择你的显卡和量化级别，查看可运行的开源模型列表

NVIDIA RTX 4090x 1 =24GB总显存(Q4 量化)

可运行 162 个开源模型

Qwen3.6 27B (Reasoning)阿里云

IQ 45.817GB

Qwen3.6 35B A3B (Reasoning)阿里云MoE

IQ 43.522GB

Qwen3.5 27B (Reasoning)阿里云

IQ 42.117GB

Gemma 4 31B (Reasoning)Google

IQ 39.218GB

Qwen3.5 27B (Non-reasoning)阿里云

IQ 37.217GB

Qwen3.5 35B A3B (Reasoning)阿里云MoE

IQ 37.122GB

Qwen3.6 27B (Non-reasoning)阿里云

IQ 37.117GB

Qwen3.5 9B (Reasoning)阿里云

IQ 32.46GB

Gemma 4 31B (Non-reasoning)Google

IQ 32.318GB

Qwen3.6 35B A3B (Non-reasoning)阿里云MoE

IQ 31.522GB

Gemma 4 26B A4B (Reasoning)GoogleMoE

IQ 31.215GB

Qwen3.5 35B A3B (Non-reasoning)阿里云MoE

IQ 30.722GB

EXAONE 4.5 33BLG AI Research

IQ 30.221GB

GLM-4.7-Flash (Reasoning)Z AI (智谱 AI)MoE

IQ 30.119GB

Nemotron Cascade 2 30B A3BNVIDIAMoE

IQ 28.419GB

Apriel-v1.5-15B-ThinkerServiceNow

IQ 28.39GB

Apriel-v1.6-15B-ThinkerServiceNow

IQ 27.69GB

Qwen3.5 9B (Non-reasoning)阿里云

IQ 27.36GB

Qwen3.5 4B (Reasoning)阿里云

IQ 27.13GB

Gemma 4 26B A4B (Non-reasoning)GoogleMoE

IQ 27.115GB

Seed-OSS-36B-Instruct字节跳动

IQ 25.222GB

Qwen3 VL 32B (Reasoning)阿里云

IQ 24.720GB

gpt-oss-20B (high)OpenAIMoE

IQ 24.513GB

NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)NVIDIAMoE

IQ 24.319GB

HyperCLOVA X SEED Think (32B)Naver

IQ 23.719GB

Qwen3.5 4B (Non-reasoning)阿里云

IQ 22.63GB

Qwen3 30B A3B 2507 (Reasoning)阿里云MoE

IQ 22.418GB

GLM-4.7-Flash (Non-reasoning)Z AI (智谱 AI)MoE

IQ 22.119GB

Nemotron 3 Nano Omni 30B A3B ReasoningNVIDIAMoE

IQ 21.418GB

gpt-oss-20B (low)OpenAIMoE

IQ 20.813GB

Qwen3 Coder 30B A3B Instruct阿里云MoE

IQ 20.018GB

Tri-21B-think PreviewTrillion Labs

IQ 20.013GB

Qwen3 VL 30B A3B (Reasoning)阿里云MoE

IQ 19.718GB

QwQ 32B阿里云

IQ 19.720GB

Devstral Small 2Mistral AI

IQ 19.514GB

Gemma 4 E4B (Reasoning)GoogleMoE

IQ 18.85GB

Tri-21B-ThinkTrillion Labs

IQ 18.613GB

Qwen3 4B 2507 (Reasoning)阿里云

IQ 18.22GB

Magistral Small 1.2Mistral AI

IQ 18.214GB

Devstral Small (May '25)Mistral AI

IQ 18.014GB

Qwen3 VL 32B Instruct阿里云

IQ 17.220GB

DeepSeek R1 Distill Qwen 32BDeepSeek

IQ 17.219GB

Magistral Small 1Mistral AI

IQ 16.814GB

Qwen3 VL 8B (Reasoning)阿里云

IQ 16.75GB

EXAONE 4.0 32B (Reasoning)LG AI Research

IQ 16.719GB

Qwen3 32B (Reasoning)阿里云

IQ 16.520GB

DeepSeek R1 0528 Qwen3 8BDeepSeek

IQ 16.45GB

Qwen3.5 2B (Reasoning)阿里云

IQ 16.31GB

Qwen3 14B (Reasoning)阿里云

IQ 16.29GB

Nanbeige4.1-3B南北阁

IQ 16.12GB

Qwen3 VL 30B A3B Instruct阿里云MoE

IQ 16.018GB

Ministral 3 14BMistral AI

IQ 16.08GB

DeepSeek R1 Distill Qwen 14BDeepSeek

IQ 15.88GB

Falcon-H1R-7BTII UAE

IQ 15.84GB

Qwen3 Omni 30B A3B (Reasoning)阿里云MoE

IQ 15.621GB

Step3 VL 10B阶跃星辰

IQ 15.56GB

Qwen3 30B A3B (Reasoning)阿里云MoE

IQ 15.318GB

QwQ 32B-Preview阿里云

IQ 15.220GB

Gemma 4 E2B (Reasoning)GoogleMoE

IQ 15.23GB

Devstral Small (Jul '25)Mistral AI

IQ 15.214GB

Mistral Small 3.2Mistral AI

IQ 15.114GB

Qwen3 30B A3B 2507 Instruct阿里云MoE

IQ 15.018GB

NVIDIA Nemotron Nano 12B v2 VL (Reasoning)NVIDIA

IQ 14.98GB

Gemma 4 E4B (Non-reasoning)GoogleMoE

IQ 14.85GB

Ministral 3 8BMistral AI

IQ 14.85GB

NVIDIA Nemotron Nano 9B V2 (Reasoning)NVIDIA

IQ 14.85GB

Qwen3.5 2B (Non-reasoning)阿里云

IQ 14.71GB

Granite 4.1 30BIBM

IQ 14.718GB

NVIDIA Nemotron 3 Nano 4BNVIDIA

IQ 14.72GB

Qwen3 32B (Non-reasoning)阿里云

IQ 14.520GB

Mistral Small 3.1Mistral AI

IQ 14.514GB

Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)NVIDIA

IQ 14.43GB

Qwen3 VL 8B Instruct阿里云

IQ 14.35GB

Qwen3 4B (Reasoning)阿里云

IQ 14.22GB

ZAYA1-8BZyphraMoE

IQ 14.15GB

Olmo 3.1 32B ThinkAllen Institute for AI

IQ 13.919GB

Qwen3 VL 4B (Reasoning)阿里云

IQ 13.73GB

Qwen2.5 Instruct 32B阿里云

IQ 13.219GB

Qwen3 8B (Reasoning)阿里云

IQ 13.25GB

NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)NVIDIAMoE

IQ 13.219GB

NVIDIA Nemotron Nano 9B V2 (Non-reasoning)NVIDIA

IQ 13.25GB

Qwen2.5 Coder Instruct 32B阿里云

IQ 12.919GB

Qwen3 4B 2507 Instruct阿里云

IQ 12.92GB

Qwen3 14B (Non-reasoning)阿里云

IQ 12.89GB

Mistral Small 3Mistral AI

IQ 12.714GB

MiniCPM-V 4.6 1.3BOpenBMB

IQ 12.71GB

Qwen3 30B A3B (Non-reasoning)阿里云MoE

IQ 12.518GB

Qwen3 4B (Non-reasoning)阿里云

IQ 12.52GB

Granite 4.1 8BIBM

IQ 12.45GB

Sarvam 30B (high)SarvamMoE

IQ 12.319GB

Olmo 3.1 32B InstructAllen Institute for AI

IQ 12.219GB

Olmo 3 32B ThinkAllen Institute for AI

IQ 12.119GB

DeepSeek R1 Distill Llama 8BDeepSeek

IQ 12.15GB

Gemma 4 E2B (Non-reasoning)GoogleMoE

IQ 12.13GB

Solar MiniUpstage

IQ 11.96GB

Llama 3.1 Instruct 8BMeta

IQ 11.85GB

EXAONE 4.0 32B (Non-reasoning)LG AI Research

IQ 11.719GB

Ministral 3 3BMistral AI

IQ 11.22GB

DeepHermes 3 - Mistral 24B Preview (Non-reasoning)Nous Research

IQ 10.914GB

Granite 4.0 H SmallIBMMoE

IQ 10.819GB

Qwen3 Omni 30B A3B Instruct阿里云MoE

IQ 10.721GB

OLMo 2 32BAllen Institute for AI

IQ 10.619GB

Qwen3 8B (Non-reasoning)阿里云

IQ 10.65GB

Qwen3.5 0.8B (Reasoning)阿里云

IQ 10.51GB

LFM2 24B A2BLiquid AIMoE

IQ 10.514GB

Phi-4Microsoft Azure

IQ 10.48GB

Gemma 3 27B InstructGoogle

IQ 10.316GB

Mistral Small (Sep '24)Mistral AI

IQ 10.213GB

Phi-3 Mini Instruct 3.8BMicrosoft Azure

IQ 10.12GB

Gemma 3n E4B Instruct Preview (May '25)GoogleMoE

IQ 10.15GB

NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)NVIDIA

IQ 10.18GB

Qwen2.5 Coder Instruct 7B 阿里云

IQ 10.05GB

Phi-4 Multimodal InstructMicrosoft Azure

IQ 10.03GB

Qwen3.5 0.8B (Non-reasoning)阿里云

IQ 9.91GB

Llama 2 Chat 7BMeta

IQ 9.74GB

Llama 3.2 Instruct 3BMeta

IQ 9.72GB

Jamba Reasoning 3BAI21 Labs

IQ 9.62GB

Qwen3 VL 4B Instruct阿里云

IQ 9.63GB

Reka Flash 3Reka AI

IQ 9.513GB

Olmo 3 7B ThinkAllen Institute for AI

IQ 9.44GB

OLMo 2 7BAllen Institute for AI

IQ 9.34GB

Molmo 7B-DAllen Institute for AI

IQ 9.25GB

Ling-mini-2.0蚂蚁 InclusionAIMoE

IQ 9.210GB

DeepSeek R1 Distill Qwen 1.5BDeepSeek

IQ 9.11GB

Gemma 3 12B InstructGoogle

IQ 8.87GB

Llama 3.2 Instruct 11B (Vision)Meta

IQ 8.77GB

DeepSeek Coder V2 Lite InstructDeepSeekMoE

IQ 8.510GB

Granite 4.1 3BIBM

IQ 8.52GB

Phi-4 Mini InstructMicrosoft Azure

IQ 8.42GB

DeepSeek LLM 67B Chat (V1)DeepSeek

IQ 8.44GB

Llama 2 Chat 13BMeta

IQ 8.48GB

Sarvam M (Reasoning)Sarvam

IQ 8.414GB

Exaone 4.0 1.2B (Reasoning)LG AI Research

IQ 8.31GB

OpenChat 3.5 (1210)OpenChat

IQ 8.34GB

Olmo 3 7B InstructAllen Institute for AI

IQ 8.14GB

Exaone 4.0 1.2B (Non-reasoning)LG AI Research

IQ 8.11GB

LFM2.5-1.2B-ThinkingLiquid AI

IQ 8.11GB

Qwen3 1.7B (Reasoning)阿里云

IQ 8.01GB

Granite 4.0 H 1BIBM

IQ 8.01GB

LFM2 2.6BLiquid AI

IQ 8.02GB

LFM2.5-1.2B-InstructLiquid AI

IQ 8.01GB

Granite 4.0 MicroIBM

IQ 7.72GB

DeepHermes 3 - Llama-3.1 8B Preview (Non-reasoning)Nous Research

IQ 7.65GB

Qwen Chat 14B阿里云

IQ 7.48GB

Command-R (Mar '24)Cohere

IQ 7.421GB

Mistral 7B InstructMistral AI

IQ 7.44GB

Molmo2-8BAllen Institute for AI

IQ 7.35GB

Granite 4.0 1BIBM

IQ 7.31GB

Granite 3.3 8B (Non-reasoning)IBM

IQ 7.05GB

LFM2 8B A1BLiquid AIMoE

IQ 7.05GB

Qwen3 1.7B (Non-reasoning)阿里云

IQ 6.81GB

Gemma 3n E4B InstructGoogleMoE

IQ 6.45GB

Llama 3 Instruct 8BMeta

IQ 6.45GB

Gemma 3 4B InstructGoogle

IQ 6.33GB

LFM2 1.2BLiquid AI

IQ 6.31GB

Llama 3.2 Instruct 1BMeta

IQ 6.31GB

LFM2.5-VL-1.6BLiquid AI

IQ 6.21GB

Apertus 8B InstructSwiss AI Initiative

IQ 5.95GB

Gemma 3 1B InstructGoogle

IQ 5.61GB

Gemma 3n E2B InstructGoogleMoE

IQ 4.84GB

Tiny Aya GlobalCohere

IQ 4.72GB

EXAONE 4.5 33B (Non-reasoning)LG AI Research

21GB

VRAM 需求为估算值（含 ~20% KV Cache 开销），实际值因推理框架（vLLM / llama.cpp / TGI）而异。 MoE 模型需全量加载权重，推理时仅激活部分参数。

开源模型选型指南

Kimi K2.6

MiMo-V2.5-Pro

DeepSeek V4 Pro (Reasoning, Max Effort)

GLM-5.1 (Reasoning)

DeepSeek V4 Pro (Reasoning, High Effort)

GLM-5 (Reasoning)

MiniMax-M2.7

MiMo-V2.5

Kimi K2.5 (Reasoning)

DeepSeek V4 Flash (Reasoning, Max Effort)

DeepSeek V4 Flash (Reasoning, High Effort)

Qwen3.6 27B (Reasoning)

Qwen3.5 397B A17B (Reasoning)

GLM-5.1 (Non-reasoning)

Qwen3.6 35B A3B (Reasoning)

Kimi K2.6 (Non-reasoning)

Qwen3.5 27B (Reasoning)

GLM-4.7 (Reasoning)

MiniMax-M2.5

Hy3-preview (Reasoning)

DeepSeek V3.2 (Reasoning)

Qwen3.5 122B A10B (Reasoning)

MiMo-V2-Flash (Feb 2026)

Kimi K2 Thinking

VRAM 兼容性检查器