LLM Inference Benchmark Explorer

DeepSeek-V4-Flash-Vision-Exp on 2× DGX Spark — FP8, vLLM, TP2, speculative decoding k=3 inference benchmark

Open this configuration in the LLM Inference Benchmark Explorer

ModelDeepSeek-V4-Flash-Vision-Exp
Parameters305B
Intelligence Index34.8
Agentic Index47.5
Device2× DGX Spark
QuantizationFP8
Max C2
Chat Capacity8
Agentic Capacity3
TP2
DP—
PP—
EnginevLLM
Speculative DecodingYes
CTTFT (ms)TPS (tok/s)Status
130035.99PASS
249623.14PASS
463316.18FAIL
89088.75FAIL

In OpenZeka's measurement, DeepSeek-V4-Flash-Vision-Exp (305B parameters), served in FP8 format with vLLM and speculative decoding on 2× DGX Spark (TP=2), reached a generation speed of 36 tok/s per request and a time to first token (TTFT) of 300 ms with a single request.

Considering the speed targets and the available KV cache capacity, the estimated capacity is 8 users for chat use and 3 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: DSpark k=3, 131K context. Model: deepseek-ai/DeepSeek-V4-Flash-Vision-Exp

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