Question 1

Can I run Gemma 4 31B on a NVIDIA GeForce RTX 5090?

Accepted Answer

Yes — NVIDIA GeForce RTX 5090 runs Gemma 4 31B fully on GPU at the Q4_K_M quantization. Full GPU inference — 32 GB VRAM meets the 24 GB recommendation.

Question 2

What quantization of Gemma 4 31B should I use on a NVIDIA GeForce RTX 5090?

Accepted Answer

For 32 GB VRAM on the NVIDIA GeForce RTX 5090, the Q4_K_M variant is the best fit. Estimated ~77.9 tokens/sec on the Q4_K_M quantization.

Question 3

How fast does Gemma 4 31B run on NVIDIA GeForce RTX 5090?

Accepted Answer

Roughly 77.9 tokens/sec for Q4_K_M. Real speed depends on context length, backend (Ollama, llama.cpp, LM Studio), and KV cache size.

Question 4

What if NVIDIA GeForce RTX 5090 is not enough for Gemma 4 31B?

Accepted Answer

Try a smaller quantization variant or run on CPU + RAM with the trade-off of slower generation.

Quantization	File Size	Min VRAM	Rec VRAM	Context	Verdict	Estimated tok/s
Q3_K_M	14.5 GB	16.5 GB	20 GB	8K / 256K	Full GPU	~85.2
Q4_K_MBest fit	18.4 GB	20.5 GB	24 GB	8K / 256K	Full GPU	~77.9
Q8_0	33.2 GB	35 GB	40 GB	8K / 256K	Hybrid CPU+GPU	~16

Can I Run Gemma 4 31B on NVIDIA GeForce RTX 5090?

Share this matchup

Every Gemma 4 31B quantization on NVIDIA GeForce RTX 5090