Can I run it?

Every model and GPU pair we check, with a verdict at the recommended quantization. Pick your card to see what runs on it, then open a pair for VRAM by quantization, context limits and speed.

Written by Jakub Rusinowski · Last updated

All cards, by model549 pairs
BitNet b1.58
DeepSeek R1
Llama 3.3
Llama 3.1 Family
Nemotron 70B
Command R Family
Phi-4 Family
Phi 3.5 Family
Qwen 2.5 Family
Gemma 2 Family
Mistral Family
Yi 1.5 Family
Granite 3.0
Llama 4
Gemma 3
Gemma 4
Llama 3.2 Family
DeepSeek V3
Qwen 3
Qwen 2.5 VL
Mistral Small 3.1
Codestral
Phi-4 Mini
StarCoder 2
OLMo 2
SmolLM2
Falcon 3
InternLM 3
Aya Expanse
Qwen3-Coder
GLM-4.7 / GLM-Z1
Aya 3B (Tiny Aya)
EXAONE 3.5
Gemma 3n
Llama 3.2 Vision
Ministral
Qwen 3.5
MiniMax M2.5
MiMo-V2-Pro
DeepSeek V3.2
Nemotron Cascade 2
GPT-OSS
Devstral
Cogito v1
Kimi K3
GLM-5 / GLM-5.1
Qwen 3.6
DeepSeek V4
Mistral Small 4
IBM Granite 4.1
MiniMax M3
Mistral Large 3
DeepSeek V4.1
Llama 4.5
Bonsai 27B
Inkling
Cosmos 3
Poolside Laguna XS 2.1
Nemotron 3 Super
GLM-5.2
MiniMax M2.7
IBM Granite 4.0
SmolLM3
VibeThinker
Magistral Small
Mistral Small 3.2
Nemotron 3 Nano Omni
Qwen3.8
MiMo-V2.5
Ministral 3
Nex-N2.5
Nex-N2
Nex-N1
GLM-5.3-Flash
North Mini Code
IBM Granite 4.2
MiniCPM-V
DeepSeek-OCR
LFM2.5
Nemotron 3.5
Ternary Bonsai 2
Muse Spark Open

By system RAM and by computer

Not everyone is asking about a graphics card. These check the same models against a system-RAM budget for CPU-only inference, or against a specific machine.