Observed arrival · 2026-10-14
Model Report Card puts AI benchmark scores under the microscope
A publication examining who ran AI benchmarks, how they were scored, and what their results actually mean.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Readers comparing AI benchmark claims and methods
- Worth noticing
- One headline says ARC-AGI-3 fell from 99.9% to 62.7% when the test harness changed.
Field notes
The index labels its term October 2026 and reports 17 filings. Its subject headings divide the archive into LLM benchmarks, agentic coding, evaluation methodology, and pricing-related coverage; individual entries show dates and source counts. The visible headlines include comparisons of conflicting lab results, a test-harness change, and a benchmark cost-per-task calculation.
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