Skip to content
localmodel.run

audio model · dia · Linux

Can I run Dia 1.6B on Nvidia DGX Spark (128GB)?

Compatibility verdict memory check
Yes, it runs fast on this GPU

Yes. Dia 1.6B runs on Nvidia DGX Spark (128GB) at fp16 (~10 GB of ~112 GB usable).

Needs ~10 GB Device usable ~112 GB

Runs at fp16 using ~10 GB of ~112 GB usable.

Peak memory
~10 GB
Usable on device
~112 GB
Device memory
128 GB
Quant
fp16
Share on X Share on Reddit

How to run it

Use PyTorch (CUDA) or nari-labs/dia at fp16. It loads as a single model and runs on a GPU.

Model dia
Type
Text to speech (dialogue)
Parameters
1.6B
Peak memory
~10 GB at fp16
License
Apache-2.0
Full Dia 1.6B 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 Dia 1.6B on other hardware

Nvidia GeForce RTX 3060 (12GB)Nvidia GeForce RTX 4070 (12GB)Nvidia GeForce RTX 5070 (12GB)Nvidia GeForce RTX 4060 Ti (16GB)Nvidia GeForce RTX 4080 (16GB)Nvidia GeForce RTX 5080 (16GB)

FAQ

Can Nvidia DGX Spark (128GB) run Dia 1.6B?

Yes. Dia 1.6B runs on Nvidia DGX Spark (128GB) at fp16 (~10 GB of ~112 GB usable).

How much memory does Dia 1.6B need?

Nvidia DGX Spark (128GB) has room to spare. At fp16 the realistic peak is ~10 GB of memory.

What do I use to run Dia 1.6B locally?

Dia 1.6B runs in PyTorch (CUDA) or nari-labs/dia. It needs a GPU.

Sources

VRAM figures are sourced peak-usage anchors at the noted quant; catalog updated 2026-11-01. See methodology.