Inkling spends 52K output tokens per task and struggles on nonstandard files

ArtificialAnlys · x · 2026-07-23

Inkling uses about 52K output tokens per AA-Briefcase task and roughly 5M tokens for the full suite.

Its weakest results are on tasks with non-standard “Other” file types rather than Excel, PowerPoint, PDF, or Word.

Related event: Inkling Evaluated on AA-Briefcase Benchmark(7 posts)→

Original post →

More from Models

Models channel →