Reaction
Theo says benchmarks undersell Opus 5.5’s value for coding
Theo calls for behavioral benchmarks, arguing Opus 5.5 remains the best value for long-running coding work.
TLDR
Artificial Analysis no longer places Opus 5.5 on its price-versus-intelligence frontier, according to Theo. But he argues the model is still the best value for long-running coding work, and calls for behavioral benchmarks to capture what he sees as a gap.
Combined views
10.7K
2 Sources, first seen ago

Comments