but they've _always_ released their stuff. Thats part of the reason why the industry uses pytorch, that and because its better than tensorflow.
In the same way that detectron and Segment anything is an industry standard.
Sure, for LLMs openAI released a product first. but its not unusual for meta to release useful models.
