Can I Run Gemma 4 on RTX 5080 Desktop (16 GB VRAM, 32 GB RAM)?

Written by Jakub Rusinowski · Last updated June 3, 2026

Yes, but it is tight

Yes, but it is tight — Gemma 4 12B (Unified) at Q8_0 needs about 14.9 GB of the 16 GB usable on RTX 5080 Desktop (16 GB VRAM, 32 GB RAM), leaving only ~1.1 GB before the runtime starts swapping. Expect ~47.9 tok/s (estimated), with room for about 8,192 tokens of context.

Confidence: low · Recommended quantization: Q8_0 · Estimated speed: ~47.9 tok/s

RTX 5080 Desktop (16 GB VRAM, 32 GB RAM) — what it gives a model

Usable memory for models	16 GB
Memory bandwidth	960 GB/s
Form factor	Desktop
Operating system	Windows or Linux
Memory upgradeable	Yes

Gemma 4 on RTX 5080 Desktop (16 GB VRAM, 32 GB RAM): memory by quantization

Quant	Memory needed	Fits 16 GB?	Max context	Est. speed	Download
F16	26.1 GB	✗ No	—	—	24 GB
Q8_0	14.9 GB	✓ Yes	8K	~47.9 tok/s	12.8 GB
Q6_K	11.9 GB	✓ Yes	32K	~59.4 tok/s	9.8 GB
Q5_K_M	10.6 GB	✓ Yes	32K	~66.8 tok/s	8.5 GB
Q4_K_M	9.4 GB	✓ Yes	32K	~75.7 tok/s	7.2 GB
Q3_K_M	7.2 GB	✓ Yes	32K	~97.5 tok/s	5.1 GB
Q2_K	6.1 GB	✓ Yes	64K	~115.8 tok/s	3.9 GB

Which Gemma 4 sizes fit

Variant	Needs at Q4_K_M	Fits?	Est. speed
Gemma 4 31B	21.3 GB	✗ Too large	—
Gemma 4 26B-A4B	18.2 GB	✗ Too large	—
Gemma 4 12B (Unified)	9.4 GB	✓ Fits	~75.7 tok/s
Gemma 4 E4B	6.8 GB	✓ Fits	~102.6 tok/s
Gemma 4 E2B	4.9 GB	✓ Fits	~138.5 tok/s

What to watch out for

Only ~1.1 GB of headroom at Q8_0: a longer context or a second application can push this into swapping.
2 larger variants of Gemma 4 do not fit and would need CPU offload or different hardware.
This model does not publish its full attention configuration, so the KV-cache share of these figures is inferred from its parameter count rather than computed exactly.

RTX 5080 desktop limitations

16 GB VRAM is the binding constraint, not compute — a slower 24 GB card runs strictly more models.

Recommended setup

Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving

How these numbers are calculated

Memory need = quantized weights + KV cache at 8,192 tokens (f16) + 0.8 GB runtime overhead.
16 GB of VRAM on the NVIDIA GeForce RTX 5080 at 960 GB/s.
32 GB of system RAM available for CPU offload when a model exceeds VRAM.
Throughput is a memory-bandwidth roofline estimate, not a measurement. It is labelled "estimated" everywhere it appears.
KV cache is inferred from the parameter count because this model does not publish its full attention configuration.

FAQ