Qwen3 32B on DGX Spark: VRAM, speed and fit
Qwen3 32B at 8-bit runs on the DGX Spark 128 GB with 40K of context at about 5.6 tokens a second.
Text generation · Chat · Ollama / LM Studio
Runs comfortably · on DGX Spark 128 GB · 8-bit (Q8_0) · Ollama / LM Studio · Chat
- GPU memory
- 50.8 GB
- of 128 GB
- Context
- 40K
- auto
- Speed
- 5.6 tok/s
- 4.2 when full
- First reply
- 2.2 s
- 2K prompt
Where the memory goes
- Model
- 32.4 GB
- Context
- 10.0 GB
- Overhead
- 713 MB
- OS
- 7.7 GB
- Free
- 77.2 GB
- Memory
- 50.8 GB of 128 GB
- Context
- 40K tokens · auto
- Speed
- 5.6 tok/s · 4.2 when full
- First reply
- 2.2 s for a 2K prompt
- Running cost
- $3.15/mo at home · $66/mo on a cloud 2× RTX A5000
Validation: Not validated
| Check | Target | Result | Status |
|---|---|---|---|
| Fits in GPU memory | 120 GB usable | 43.1 GB needed | Pass |
| Headroom | 5% free | 77.2 GB free (64%) | Pass |
| Context for chat | 16K or more | 40K | Pass |
| Writing speed | 15 tok/s or more | 5.6 tok/s | Fails |
| First reply | under 3 s, 2K prompt | 2.2 s | Pass |
Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.
The setup
- Model
- Qwen3 32B
- Parameters
- 32.8B
- Precision
- 8-bit (Q8_0)
- Context
- 40K, auto
- Chats at once
- 1
- Use
- Chat
- Runs in
- Ollama / LM Studio
- Hardware
- DGX Spark 128 GB
- GPU can use
- 120 GB of 128 GB
Running cost
- At home
- $3.15/mo
- 144 W at the wall
- Renting instead
- $66/mo
- 2× RTX A5000 at $0.54/h
4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 21, 2026.
How to run it
- llama.cpp
llama-server -hf Qwen/Qwen3-32B-GGUF- vLLM
vllm serve Qwen/Qwen3-32B- MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3-32B-4bit- Model card
- huggingface.co/Qwen/Qwen3-32B
Qwen3 32B on similar machines
- M5 Ultra 512 GBRuns
- Ryzen AI Max+ 395 64 GBRuns
- Ryzen AI Max+ 395 128 GBRuns
- RTX SparkRuns
- Dual DGX SparkRuns
- DGX Station GB300Runs
- 2× RTX PRO 6000 BlackwellRuns
- 4× RTX PRO 6000 BlackwellRuns
All machines: Qwen3 32B VRAM requirements. Everything that runs here: what can the DGX Spark run?
Other models that run on DGX Spark
Questions
Can Qwen3 32B run on DGX Spark?
Qwen3 32B at 8-bit runs on the DGX Spark 128 GB with 40K of context at about 5.6 tokens a second.
Report VR-1C634TL, calculated Nov 22, 2026 from published specs, not benchmarked.