Hacker News
Calvin French-Owen Puts a Price on the Small-Model Shift: $0.10 a Day Where Sonnet-Class Cost $1
French-Owen's August 26 post argues small models crossed the viability line for consumer apps, using his own daily news personalization eval as the measurement: about $0.10 a run on gpt-5.6-luna at roughly 100 tokens per second, against about $1 on a Sonnet-class model. He names GLM 5.3 as a new point on the efficiency frontier and Fable 5 and 5.6 Sol as the expensive capable tier. His framing for builders is that 95% of business work is responsive routine execution rather than novel problem-solving, so the fast-cheap-good-enough tier addresses most of the actual demand.
↳ Follow the thread