Sonnet 5.5's fastest setup used subagents; Opus 5.5's used a five-agent team in one ProgramBench evaluation
A user reports that the two models ran fastest with different setups in a multi-agent evaluation on ProgramBench.
TLDR
A user shared results from a ProgramBench multi-agent evaluation: Sonnet 5.5's fastest setup used subagents, while Opus 5.5 ran fastest with a five-agent team.
Combined views
1.5K
1 Source, first seen 4h ago
23 likes
