Small Models Have Arrived: GPT-5.6 Luna Cuts AI Costs to Cents

Calvin French-Owen has been testing OpenAI's GPT-5.6 Luna and finds it shockingly capable, fast, and cheap—running complex research threads for tens of cents. He argues that the plummeting cost of small, fast models unlocks consumer AI apps, which have been held back by token costs. For businesses, he distinguishes between "IQ 180" work (rare, genius solutions) and "token spewer" work (responsive, multi-front execution), noting that 95% of a founder's time is the latter. He predicts demand for fast/cheap/good-enough models is about to take off.