Most providers don’t publish a real number for what you get. Anthropic says “5× Free per 5-hour session”, Google says “4× baseline”, xAI says “a shared weekly pool”. Where nobody publishes a figure, we say so instead of inventing one.
Four real jobs. The pick, the band, and the exams nobody sat.re-scored at API default
Ship a feature in a legacy codebase
Evidence favoursGemini 3.6 Flashclear
Marks · Agentic coding
epoch-swe-bench-verified
LMArena WebDev Arena1527.83at high
tau2-airline
tau2-retail
Read 200 pages and argue with it
Not measurednothing to weigh
WhyNo benchmark is attributed to this job yet, so nothing has been weighed and we hold no opinion.
Marks · Long-document comprehension
epoch-fictionlivebench
Make a pitch deck not look like a pitch deck
Evidence favoursGemini 3 Flash Previewclear
Marks · Multimodal
LMArena Vision Arena1337.46at high
Name my band
Evidence favoursGLM-5.3-Flashclear
Marks · Writing & editing
LMArena Text Arena1471.89at max
Your job
Marks
—
—
—
—
You pay for six AIs. You use one.
A slashed box is not a bad score. It is an exam nobody sat, and what is slashed above is the most useful thing on this page.