TECH Signal 498
Small models like gpt-5.6-luna cut task costs from ~$1 to ~$0.10
Small AI models have crossed a cost threshold where tasks that previously cost ~$1 with Sonnet-class models now run for ~$0.10, making consumer AI business models economically viable.
Token costs have been the structural barrier preventing consumer AI companies from following the traditional build-cheap-attract-users-monetize playbook. At ~$0.10 per complex task, the unit economics shift from untenable to plausible for subscription products.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
gpt-5.6-luna achieves ~100 tps and handles coding, email, and knowledge base tasks with API costs in the tens of cents
A personalized news eval costing ~$1 with Sonnet-class models now costs ~$0.10 with luna
An estimated ~95% of business operational work is fast/cheap/good-enough 'token spewer' work rather than frontier-level problem solving
THE CLUSTER