Announcement
DeepSeek V4.1 Flash joins AA-AgentPerf; Nvidia GB300 leads throughput per MW at 100 tokens/s per user
Artificial Analysis reports GB300 reached up to 261 million tokens/s per MW, ahead of B300's roughly 220 million.
TLDR
Artificial Analysis added DeepSeek V4.1 Flash to AA-AgentPerf, its benchmark of agentic coding sessions with prompt caching and tool-call delays. At 100 tokens/s per user, it reports GB300 reached up to 261 million tokens/s per MW, versus about 220 million for B300 and 128 million for AMD's MI355X. It says V4.1 Flash scores 39 on its Intelligence Index, versus 36 for DeepSeek V4 Pro 0813, with about a third of the parameters.
Combined views
3.5K
6 Sources, first seen ago
