text model · Sarvam · Linux
Can I run Sarvam-30B on Nvidia DGX Spark (128GB)?
Yes. Sarvam-30B runs on Nvidia DGX Spark (128GB) at Q4_K_M (~21.7 GB of ~112 GB usable).
Runs at Q4_K_M using ~21.7 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 ~90.3 GB of headroom, room to step up to FP16 for higher quality.
- Q4_K_M needed
- ~21.7 GB
- Usable on device
- ~112 GB
- Device memory
- 128 GB
- Best quant
- Q4_K_M
Which quant fits
Run it
Pick your tool. All 2 load the same Q4_K_M weights.
llama-cli -hf sarvamai/sarvam-30b-gguf:Q4_K_M lms get sarvamai/sarvam-30b-gguf How to run it
On Linux use Ollama (One-line install, simple single-user UX.).
- Parameters
- 30B (MoE, 2.4B active)
- Q4_K_M size
- 19.6 GB
- Context
- 64k
- 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 Sarvam-30B on other hardware
FAQ
Can Nvidia DGX Spark (128GB) run Sarvam-30B?
Yes. Sarvam-30B runs on Nvidia DGX Spark (128GB) at Q4_K_M (~21.7 GB of ~112 GB usable).
How much memory does Sarvam-30B need?
Nvidia DGX Spark (128GB) has room to spare. At Q4_K_M the weights are ~19.6 GB; with KV cache and runtime overhead, budget ~21.7 GB at a 4k context. It is a Mixture-of-Experts model (30B total / 2.4B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run Sarvam-30B 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/sarvam-30b/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 · 1 source
-
sarvam.ai · 1 source
-
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.