Qwen3 Coder 30B-A3B Instruct › NVIDIA DGX Spark (GB10)

Qwen3 Coder 30B-A3B Instruct on the NVIDIA DGX Spark (GB10)

Best measured: 61.1 tokens/s generation and 2,987 tokens/s prompt processing (Q8_0, CUDA, FA on).

For comparison, our size-scaling estimate for this combination is 36.7 tok/s.

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
Q8_0	CUDA FA on	61.1	2,987	tg32	ggml-org maintainers Benchmarked repo: https://huggingface.co/ggml-org/Qwen3-Coder-30B-A3B-Instruct-Q8_0-GGUF	llama.cpp benches/ (maintainer runs)

tg128 = generating 128 tokens; pp512 = reading a 512-token prompt (llama-bench). LocalScore averages several prompt and output lengths, so its numbers run lower than tg128.

Check prices

Amazon Newegg

As an Amazon Associate I earn from qualifying purchases.

No card yet?

Try it on a rented GPU (RunPod)

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.

All results for the NVIDIA DGX Spark (GB10) · Qwen3 Coder 30B-A3B Instruct on other hardware