Not definitive, since distillation exists, but on speechmap word frequency it looks most like glm-5.2

i don't have a problem with releasing a compressed version of a model, but the presentation seems extremely dishonest. https://t.co/QsGMzQEfEK