
GoodMem · Benchmark
Memory That Pays for Itself
Controlled A/B benchmark — GoodMem's memory layer cut aggregate token use 28% and reasoning iterations 23%, at no measurable loss of answer quality.
- token burn
- −28%
- reasoning steps
- −23%
- quality change
- ≈ 0

