Techstrong.ai posted about Signal65's PINNACLE benchmark. The benchmark assesses AI platforms on whole job completion, cost per correct task, and sustained platform capacity. It moves the argument from abstract intelligence and token speed to a harder enterprise measure, according to the linked source on Techstrong.ai. The post states early results show a median 28-point drop when messy enterprise data enters the workflow, with only Claude Opus 5 and GPT-5.6 Sol staying competitive.