Artificial Analysis: DeepSeek V4.1 Flash on GB300 NVL72 is up to 1.5x faster per agent in a month

Artificial Analysis says DeepSeek V4.1 Flash speeds on NVIDIA GB300 NVL72 have risen by up to 1.5x tokens per second per agent over the past month on AA-AgentPerf, its benchmark of datacenter systems running realistic agentic coding workloads.

• Speed: 848 tok/s for a single agent, up from 596 (1.42x).
• Throughput: about 1.6x the throughput per GPU at about 80 tok/s per agent.

Artificial Analysis attributes the gain to a newer vLLM version, new vLLM kernels for DeepSeek V4.1 on GB300, FlashInfer autotuning, fuller CUDA graph coverage, tuned speculative decoding, and running prefill and decode on the same GPUs in NVIDIA's updated Dynamo + vLLM setup.