Quick benchmark: Sol 6.1 inference is nearly 6x slower than Opus despite better token efficiency

RexDouglass · x · 2026-10-01

Gael Breton ran a quick test showing Sol 6.1 inference is almost 6x slower than Opus. Because the model is far more token-efficient, it doesn't feel 6x slower in practice, but he argues slow inference turns a strength into a weakness and hopes OpenAI fixes it. Otherwise he rates the model very positively.

Related event: Sol 6.1 Tests Show Slower Inference but Better Token Efficiency(2 posts)→

Original post →

More from Models

Models channel →