Skip to content

Can RTX 3090 run Gemma 3 27B IT?

Comfortable: GGUF Q4_K_M needs 18.3 GB at 4K, batch 1; headroom 5.7 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_M18.3ComfortableVerified file
GGUF Q3_K_M15.1ComfortableVerified file
AWQ INT415.1ComfortableEstimated weights

Context capacity · GGUF Q4_K_M

ContextTotal GBVerdict
4K18.3Comfortable
16K19.3Comfortable
32K20.7Comfortable
128K29.0Does not fit

Generation speed

GGUF Q4_K_M: 36–48 tok/s · Derived range

Other model / GPU pairs