can-i-run-this-llm

Can I run this LLM?

Can a NVIDIA RTX 5090 (32GB) run Gemma 4 31B?

Gemma · Apache 2.0 · estimated for a 32.0 GB machine at 4K context.

Yes — it runs great.

Best fit: Gemma 4 31B at Q5_K_M · 21.7 GB · 50–75 tok/s on a NVIDIA RTX 5090 (32GB).

Every quantization on this machine

QuantSizeSpeedFit
Q3_K_M14.7 GB68–102 tok/sRuns great
Q4_K_M18.3 GB57–86 tok/sRuns great
Q5_K_M21.7 GB50–75 tok/sRuns great
Q6_K25.2 GB44–66 tok/sTight / slow
Q8_032.6 GB35–53 tok/sWon't fit
FP1661.4 GB20–30 tok/sWon't fit

Ran Gemma 4 31B on a NVIDIA RTX 5090 (32GB)? Add your real numbers →