text model · gpt-oss · Linux
Can I run gpt-oss 20B on Nvidia DGX Spark (128GB)?
Yes. gpt-oss 20B runs on Nvidia DGX Spark (128GB) at Q4_K_M (~13.2 GB of ~112 GB usable).
Runs at Q4_K_M using ~13.2 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 ~98.8 GB of headroom, room to step up to FP16 for higher quality.
- Q4_K_M needed
- ~13.2 GB
- Usable on device
- ~112 GB
- Device memory
- 128 GB
- Best quant
- Q4_K_M
Which quant fits
Run it
Pick your tool. All 3 load the same Q4_K_M weights.
ollama run gpt-oss:20b llama-cli -hf ggml-org/gpt-oss-20b-GGUF:Q4_K_M lms get ggml-org/gpt-oss-20b-GGUF - Parameters
- 21B (MoE, 3.6B active)
- Q4_K_M size
- 11.28 GB
- Q8_0 size
- 0.86 GB
- Context
- 128k
- Ollama tag
- gpt-oss:20b
- Memory
- 128 GB unified
- Usable for weights
- ~112 GB
- Power draw
- ~140 W
- Best runtime
- llama.cpp (CUDA) / Ollama (CUDA)
You could also run
Run gpt-oss 20B on other hardware
FAQ
Can Nvidia DGX Spark (128GB) run gpt-oss 20B?
Yes. gpt-oss 20B runs on Nvidia DGX Spark (128GB) at Q4_K_M (~13.2 GB of ~112 GB usable).
How much memory does gpt-oss 20B need?
Nvidia DGX Spark (128GB) has room to spare. At Q4_K_M the weights are ~11.28 GB; with KV cache and runtime overhead, budget ~13.2 GB at a 4k context. It is a Mixture-of-Experts model (21B total / 3.6B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run gpt-oss 20B 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
A live badge for your model card or README, updated as the data is. [](https://localmodel.run/can-i-run/gpt-oss-20b/nvidia-dgx-spark-128gb) Sources
-
docs.nvidia.com · 2 sources
-
github.com · 1 source
-
huggingface.co · 2 sources
-
nvidia.com · 1 source
-
nvidianews.nvidia.com · 1 source
-
ollama.com · 3 sources
-
tomshardware.com · 1 source
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.