Ultrafast mode is +1,300% faster on GPT-5.6 Sol. Insane. I made a post last week about how we're entering the era of AI cost efficiency and speed. For the past year, AI labs have been primarily focused on pure intelligence. I have no doubt that LLM intelligence will continue to make huge leaps, but what's becoming increasingly important is cost/speed. The models are good enough for 95% of daily use cases, but they're either a) too slow or b) too expensive These problems should be solved soon.(Miles Deutscher)