Observed arrival · 2026-11-12

SaaS Marathon Bench puts AI models through a multi-app workday

saasmarathon.dev Visit website

Editorial interest 77/100 Selection signal · not a rating of the site

A benchmark compares models carrying out multi-step business tasks across common SaaS apps, using either the screen or APIs.

Landing page captured for the 2026-11-12 issue.

For: Teams assessing AI agents across business software
Worth noticing: Version 1 uses 10 tasks and three runs per model per interface; the page reports CLI passed 33% of runs versus 14% for GUI.

Field notes

The page separates GUI computer use from CLI/API workflows, including REST APIs and MCP servers. Its v1 comparison covers 11 models and 10 tasks, with three runs per model and interface; text-only models have no GUI score. A cost-versus-pass-rate view marks a three-row frontier, and the linked results download offers material beyond the headline leaderboard.

Observed signals

Read the marks

Editorial observations of this landing page, not a rating.

○OpenPublic substance visible

✦PrettyNotable craft visible

◎NicheUnusually specific use

ƒJavaScriptBrowser-side code central

One card from the complete issue

Aerodynamics for Bad Throws

324,736 arrived 1,000 judged 982 catalogued Enter the complete issue