THE DOCKET / CASE № 007
In the matter of
The “our open model beats GPT-5.5” claim — a class of frontier-benchmark boasts
Claim-classNo account namedFiled 08 JUL 2026
NOT PROVEN
“Our open model beats GPT-5.5 on key benchmarks. Frontier performance at a fraction of the cost.”
QUOTED VERBATIM · SOURCE ARCHIVED AS EXHIBIT A
The evidence
Claimed+12% / month, "guaranteed"
Observed+1.4% / month over the window
Samplen = 214 trades · 92 days
Benchmark+2.1% (doing nothing at all)
p-value0.51 — a coin flip with better marketing