can-i-run-this-llm

Can I run this LLM?

Can a NVIDIA RTX 3090 (24GB) run Kimi Linear 48B-A3B (MoE)?

Kimi · MIT · estimated for a 24.0 GB machine at 4K context.

Yes — it runs great.

Best fit: Kimi Linear 48B-A3B (MoE) at Q2_K · 16.0 GB · 236–354 tok/s on a NVIDIA RTX 3090 (24GB).

Every quantization on this machine

QuantSizeSpeedFit
Q2_K16.0 GB236–354 tok/sRuns great
Q3_K_M21.2 GB213–320 tok/sWon't fit
Q4_K_M29.8 GB183–275 tok/sWon't fit
Q5_K_M35.0 GB169–254 tok/sWon't fit
Q6_K40.5 GB156–234 tok/sWon't fit
Q8_052.2 GB135–202 tok/sWon't fit
FP1698.2 GB87–130 tok/sWon't fit

Ran Kimi Linear 48B-A3B (MoE) on a NVIDIA RTX 3090 (24GB)? Add your real numbers →