Cost–reliability frontier by model release date
Models appear on their release dates in cost–recall space. A navy step curve traces the Pareto frontier; models fade to grey the moment a newer release dominates them.
Verifier Cost–Reliability Frontier · APE-CRED v0.92
APR 09, 2024
cost–recall frontier
frontier at APE start
single-shot model
agentic model + harness
PROJECT APE
·
ape.socialcatalystlab.org
00:00 / 01:11
Speed
0.5×
1×
1.5×
2×