LLM Inference Benchmark Explorer

Qwen3-Coder-30B-A3B-Instruct on 1× DGX Spark — NVFP4, vLLM, TP1 inference benchmark

Open this configuration in the LLM Inference Benchmark Explorer

ModelQwen3-Coder-30B-A3B-Instruct
Parameters30B
Intelligence Index9.6
Agentic Index—
Device1× DGX Spark
QuantizationNVFP4
Max C32+
Chat Capacity99
Agentic Capacity24
TP—
DP—
PP—
EnginevLLM
Speculative Decoding—
CTTFT (ms)TPS (tok/s)Status
16666.76PASS
24955.34PASS
45942.97PASS
87634.04PASS
167827.62PASS
329222.03PASS

In this measurement, Qwen3-Coder-30B-A3B-Instruct (30B parameters), served in NVFP4 format with vLLM on 1× DGX Spark, reached a generation speed of 66.8 tok/s per request and a time to first token (TTFT) of 66 ms with a single request.

Considering the speed targets and the available KV cache capacity, the estimated capacity is 99 users for chat use and 24 for agentic use. The realistic capacity will likely fall between these two values. In scenarios dominated by coding, tool use, long workflows and multi-agent use, capacity approaches the agentic estimate; where shorter interactions, standard conversations and lighter tasks dominate, it approaches the chat estimate.

Notes: NVFP4 is 2.2x faster than FP16 at C1

Intelligence Index and Agentic Index values are published by Artificial Analysis and are reproduced here with attribution.