Skip to content
Trending storyDeveloping

OpenBMB highlights 370 tokens/sec for MiniCPM-V 4.7 with DSPARK drafter

1 article1 sourcesince Oct 15Last article 54m ago ·

Overview

AISummary of 1 article

OpenBMB says MiniCPM V4.7 35B A3B, paired with a DSPARK speculative-decoding drafter, reached 370 tokens/sec in a community test, as reported by Tech2Wild.

In the 10-second run the setup produced 3,589 tokens, with a 0.30-second average time to first token and a 10-second average end-to-end latency. OpenBMB's post on X calls the result a community achievement in speculative decoding and says it is glad to see developers push that technique's limits. The only source reported is the test as relayed by Tech2Wild and OpenBMB's post, with no independent benchmark published.

Written by AI from the articles below · updated Oct 15, 5:10 AM ET

Check the sources:

Latest developmentOct 15, 4:59 AM ET
MiniCPM-V 4.7 hits 370 tokens/sec with DSPARK speculative decoding

Article timeline

The articles in this story. Times are ET.

Oct 15
  1. OpenBMBOfficial
    MiniCPM-V 4.7 hits 370 tokens/sec with DSPARK speculative decoding

    AITech2Wild reports that OpenBMB's MiniCPM v4.7 35B A3B, paired with a DSPARK drafter, sustained 370.0 tokens/sec aggregate in a 10-second test. The run produced 3,589 tokens with a 0.30-second average time to first token and a 10-second average end-to-end latency. OpenBMB's post celebrates the result as a community speculative decoding achievement.

Heat trend

Not enough continuous observations to show a trend yet.