Skip to content
localmodel.run

text model · Gemma · Linux

Can I run Gemma 4 31B on Nvidia DGX Spark (128GB)?

Compatibility verdict VRAM threshold engine
Yes, it runs GPU accelerated ~10 tok/s est.

Yes. Gemma 4 31B runs on Nvidia DGX Spark (128GB) at Q4_K_M (~20.5 GB of ~112 GB usable).

Needs ~20.5 GB Device usable ~112 GB

Runs at Q4_K_M using ~20.5 GB of ~112 GB usable. You have room for FP16 for higher quality.

That figure is at a 4k context and moves about ±15% as context length changes. Nvidia DGX Spark (128GB) leaves ~91.5 GB of headroom, room to step up to FP16 for higher quality.

Q4_K_M needed
~20.5 GB
Usable on device
~112 GB
Device memory
128 GB
Best quant
Q4_K_M
Share on X Share on Reddit

Which quant fits

Quant ladder vs ~112 GB usable
Q2_K
~15.9 GB
Q3_K_M
~18.2 GB
Q4_K_M
~20.5 GB
Q5_K_M
~25.5 GB
Q6_K
~29 GB
Q8_0
~34.8 GB
FP16
~63.6 GB
The line marks Nvidia DGX Spark (128GB)'s ~112 GB budget; rungs past it are too large.

Running cost · estimate

Power & economics est.
Power draw
~140 W
Electricity / 1M tokens
~$0.58

At ~$0.15/kWh and the estimated ~10 tok/s, a million generated tokens costs about $0.58 in electricity. TDP is the peak draw, so this is an upper bound. Assumptions.

Run it

Install commands Linux

Pick your tool. All 3 load the same Q4_K_M weights.

Ollama
$ ollama run gemma4:31b
llama.cpp
$ llama-cli -hf unsloth/gemma-4-31B-it-GGUF:Q4_K_M
LM Studio
$ lms get unsloth/gemma-4-31B-it-GGUF
Model Gemma
Parameters
32.7B
Q4_K_M size
18.32 GB
Q8_0 size
32.64 GB
Context
256k
Ollama tag
gemma4:31b

Set up this model at 4K →

Full Gemma 4 31B requirements →
Device Linux
Memory
128 GB unified
Usable for weights
~112 GB
Power draw
~140 W
Best runtime
llama.cpp (CUDA) / Ollama (CUDA)
Best models for Nvidia DGX Spark (128GB) →

You could also run

Run Gemma 4 31B on other hardware

Nvidia GeForce RTX 4090 (24GB)Nvidia GeForce RTX 3090 (24GB)AMD Radeon RX 7900 XTX (24GB)Nvidia GeForce RTX 5090 (32GB)32GB RAM Laptop (CPU/iGPU only)Apple M5 (32GB)

FAQ

Can Nvidia DGX Spark (128GB) run Gemma 4 31B?

Yes. Gemma 4 31B runs on Nvidia DGX Spark (128GB) at Q4_K_M (~20.5 GB of ~112 GB usable).

How much memory does Gemma 4 31B need?

Nvidia DGX Spark (128GB) has room to spare. At Q4_K_M the weights are ~18.32 GB; with KV cache and runtime overhead, budget ~20.5 GB at a 4k context.

What is the best tool to run Gemma 4 31B on Linux?

Ollama for a simple setup; vLLM for the most speed. vLLM shines for multi-user serving/throughput. For a single local chat, Ollama or llama.cpp is simpler and lighter.

Embed this

Gemma 4 31B on Nvidia DGX Spark (128GB) compatibility badge A live badge for your model card or README, updated as the data is.
Markdown
$ [![Gemma 4 31B on Nvidia DGX Spark (128GB)](https://localmodel.run/badge/gemma-4-31b/nvidia-dgx-spark-128gb.svg)](https://localmodel.run/can-i-run/gemma-4-31b/nvidia-dgx-spark-128gb)

Sources

Weights are measured from GGUF files; KV cache and overhead are computed, so totals can vary ~15% with context and runtime. Any tok/s is a bandwidth estimate. See methodology.