RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can gemma 4 31B it run on the 2x RTX 5090?

Yes. At 8K context the best fit is Q8_0, using about 33.3 GB of the 60.8 GB usable on the 2x RTX 5090. Estimated speed ceiling: ~55 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K)13.2 GB Fits13.9 GB Fits17.6 GB Fits32.6 GB Fits153
3-bit (Q3_K_M) •16.0 GB Fits16.7 GB Fits20.4 GB Fits35.4 GB Fits122
4-bit (Q4_K_M) •19.4 GB Fits20.0 GB Fits23.7 GB Fits38.7 GB Fits98
5-bit (Q5_K_M) •22.5 GB Fits23.1 GB Fits26.9 GB Fits41.9 GB Fits83
6-bit (Q6_K) •25.8 GB Fits26.4 GB Fits30.2 GB Fits45.2 GB Fits71
8-bit (Q8_0) •32.7 GB Fits33.3 GB Fits37.1 GB Fits52.1 GB Fits55
16-bit (F16/BF16) •59.5 GB Tight60.1 GB Tight63.9 GB No78.9 GB No29

• real GGUF file size from unsloth/gemma-4-31B-it-GGUF. Speed is a bandwidth ceiling (1792 GB/s), not a benchmark. Methodology.

More

gemma 4 31B it on every GPU and Mac · Every model that runs on the 2x RTX 5090 · Try other settings in the calculator

Data: Hugging Face Hub (google/gemma-4-31B-it), NVIDIA specs. Data as of 2026-11-17.