Skip to content
Trending storyDeveloping

Fable averages 1.60 tool calls per turn on VIALS-AA, highest on leaderboard

2 articles1 sourcesince Nov 9Last article 46m ago ·

Overview

AISummary of 2 articles

Artificial Analysis reports that Fable averages 1.60 tool calls per turn across the VIALS-AA benchmark, the highest rate of any model on the leaderboard.

The figure comes from the updated VIALS-AA results the firm published on the same day, which cover scores, cost per task and tool use.

The results are on Artificial Analysis's evaluations page, where model trajectories can be explored. The reports do not give Fable's scores or cost per task.

Written by AI from the articles below · updated Nov 9, 10:11 PM ET

Check the sources:

Latest developmentNov 9, 10:07 PM ET
Fable averages 1.60 tool calls per turn on VIALS-AA, highest on leaderboard

Developments

2 developments

  1. Nov 9, 10:07 PM ET · 1 article
    Artificial Analysis reports Fable's tool-call rate on VIALS-AA
    Artificial Analysis: Fable averages 1.60 tool calls per turn on VIALS-AA, highest on leaderboard
  2. Nov 9, 10:07 PM ET · 1 article
    Artificial Analysis publishes updated VIALS-AA results
    Artificial Analysis: Artificial Analysis publishes updated VIALS-AA results with scores, cost and tool use

Article timeline

The articles in this story. Times are ET.

Nov 9
  1. Artificial AnalysisOfficial
    Fable averages 1.60 tool calls per turn on VIALS-AA, highest on leaderboard

    AIArtificial Analysis reports that Fable averages 1.60 tool calls per turn across the VIALS-AA benchmark. That is the highest rate of any model on the leaderboard.

    Image from @ArtificialAnlys's post
  2. Artificial AnalysisOfficial
    Artificial Analysis publishes updated VIALS-AA results with scores, cost and tool use

    AIArtificial Analysis has published updated VIALS-AA evaluation results, covering scores, cost per task and tool use. The post links to a launch article and an evaluations page for model trajectories.

Heat Trend

Not enough continuous observations to show a trend yet.