Skip to content

Can RTX 4090 run Gemma 4 31B IT?

Comfortable: GGUF Q4_0 needs 20.1 GB at 4K, batch 1; headroom 3.9 GB.

Text input, FP16 KV cache, one card. Unified-memory capacity uses the GPU guide’s usable limit; negative headroom may require host offload. Speed source settings may differ from this 4K estimate.

Measured runs

No measured run for this pair. Sizing is an estimate, not a runtime test.

Quant memory at 4K

QuantTotal GBVerdictWeight sizing
GGUF Q4_K_M22.1TightVerified file
GGUF Q4_020.1ComfortableVerified file
GGUF Q8_035.7Does not fitVerified file
GGUF Q3_K_M18.3ComfortableVerified file
GGUF Q5_K_M25.3Does not fitVerified file

Context capacity · GGUF Q4_0

ContextTotal GBVerdict
4K20.1Comfortable
16K21.1Comfortable
32K22.5Tight
128K30.7Does not fit

Generation speed

GGUF Q4_0: 40–53 tok/s · Derived range

Other model / GPU pairs