We raised $4m to help teams build together with AI
Get started for free

Type.com benchmarks

Measured comparisons of Type.com against other ways to run AI automations. Every report publishes its method, per-task results and raw data.

Updated

Reports

Published benchmarks

  • Version 1 · November 2026 · AutomationBench · 250 tasks · 750 paired runs · Claude Opus 5.5

    Recurring automation benchmark

    Type.com vs Claude Code (Opus 5.5)

    Repeat runs were 30× faster on 97% less model spend than Claude Code, with no loss in accuracy.

    faster repeat runs
    30×
    less model spend
    97%

What every report includes

Every benchmark shows its work

A benchmark number means little without the method behind it. Each report on this page includes:

  • The method: tasks, model, limits and how runs were graded
  • Results for every setup we compared, including cost and run time
  • Results for every task, not only the averages
  • Where our approach failed, not only where it won
  • The raw data, as JSON and CSV
  • A version number and the date of the results

Run your repeat work 30× faster.

Describe the job once. Type.com writes the code, calls Claude when a step needs judgment, and runs it on your team's connections.

Get started for free