LLM Hardware Advisor

loading…

Your Hardware

GPU Model

Select GPU to auto-fill VRAM & enable performance estimates

GPU VRAM 12 GB

GPU memory for model + KV cache

System RAM 32 GB

RAM for offload / CPU-only layers

RAM Type DDR4 · 50 GB/s

Used for partial GPU offload speed estimates

Min. Context 8 192

Filter: models must support at least this context window

Perf. Context 4 096

Actual context you use — affects KV cache size & speed estimate

Fits fully on GPU

Model on GPU, KV overflows to RAM

Partial GPU offload (slower)

Doesn't fit at all

Reasoning9%

Math9%

Code9%

Search9%

Communication9%

Vision8%

Tool Calling9%

Long Context9%

Finance8%

Legal8%

Healthcare8%

Knowledge5%

Total: 100 — weights auto-normalized when computing scores

Hardware: