Skip to content
Trending storyDeveloping

Mistral's Le Chonk scores 42% on Dense 200 visual grounding benchmark

1 article1 sourcesince Oct 12Last article 1h ago ·

Overview

AISummary of 1 article

Mistral AI says its Le Chonk model scores 42% on the Dense 200 visual grounding benchmark, which tests how precisely a model locates objects in an image.

In a post on X, the company says the score is ahead of frontier closed models. The post gives no comparison figures for those models, and no independent evaluation of the result has been reported.

Written by AI from the articles below · updated Oct 12, 1:05 PM ET

Check the sources:

Article timeline

The articles in this story. Times are ET.

Oct 12
  1. Mistral AIOfficial
    Mistral's Le Chonk scores 42% on Dense 200 visual grounding benchmark

    AIMistral AI says its Le Chonk model pinpoints exactly where objects are in an image. It scores 42% on the Dense 200 benchmark, which the post says is ahead of even frontier closed models.

    Video from @MistralAI's post

Heat trend

Not enough continuous observations to show a trend yet.