🤖 Air-Gapped AI

Offline Model Selection Guide

Enter your hardware constraints and security task requirements. Get a specific model and quantisation recommendation with full reasoning.

⚙️ Your Environment & Requirements

📊 Full Model Reference — Air-Gapped Security Use

ModelSizeMin VRAM (Q4)Min VRAM (Q8) Log AnalysisAlert TriageTool CallingJSON OutputContextBest For
Phi-3 Mini 3.8B3.8B2.5 GB4 GB★★★★★☆PoorOK128KCPU-only, resource-constrained
Mistral 7B v0.37B4.5 GB8 GB★★★★★★★LimitedOK32KSingle GPU, fast inference
Llama 3.1 8B8B5 GB9 GB★★★★★★★LimitedGood128KBest 8B overall — long context
Phi-3 Medium 14B14B9 GB15 GB★★★★★★★★LimitedGood128KBest quality per VRAM on 16-24GB
Mistral Nemo 12B12B8 GB13 GB★★★★★★★LimitedGood128KEfficient, good instruction following
Qwen 2.5 7B7B4.5 GB8 GB★★★★★★★LimitedGood128KStrong instruction following at 7B
Llama 3.1 70B70B43 GB75 GB★★★★★★★★★★GoodGood128KBest overall air-gapped quality
Qwen 2.5 72B72B45 GB76 GB★★★★★★★★★★GoodExcellent128KBest JSON/structured output at 70B
Mistral Large 2 123B123B70 GB130 GB★★★★★★★★★★ExcellentExcellent128KTop tier — requires significant hardware