Skip to content
Trending storyDeveloping

Artificial Analysis benchmarks DeepSeek V4.1 Flash on NVIDIA and AMD hardware in AA-AgentPerf

1 article1 sourcesince Oct 14Last article 1h ago ·

Overview

AISummary of 1 article

Artificial Analysis has added DeepSeek V4.1 Flash to its AA-AgentPerf benchmark, which replays agentic coding sessions and measures tokens per second per megawatt.

At 100 tokens/s per user, the post reports NVIDIA GB300 at about 261M tokens/s per MW, NVIDIA B300 at about 220M, and AMD MI355X at about 128M.

The same post says V4.1 Flash scores 39 on the Artificial Analysis Intelligence Index, above V4 Pro 0813 at 36, while using about a third of its parameters. These are the firm's own figures, reported in a single X post.

Written by AI from the articles below · updated Oct 14, 3:13 AM ET

Check the sources:

Latest developmentOct 14, 3:04 AM ET
Artificial Analysis benchmarks DeepSeek V4.1 Flash across NVIDIA and AMD hardware

Article timeline

The articles in this story. Times are ET.

Oct 14
  1. Artificial AnalysisOfficialPick
    Artificial Analysis benchmarks DeepSeek V4.1 Flash across NVIDIA and AMD hardware

    AIArtificial Analysis added DeepSeek V4.1 Flash to its AA-AgentPerf benchmark, which replays agentic coding sessions and measures throughput per megawatt. At 100 tokens/s per user, NVIDIA GB300 leads with about 261M tokens/s per MW, ahead of B300 at about 220M and MI355X at about 128M. The post also reports that V4.1 Flash scores 39 on the Artificial Analysis Intelligence Index, above V4 Pro 0813 at 36, with about a third of its parameters.

    Image from @ArtificialAnlys's post

Heat trend

Not enough continuous observations to show a trend yet.