Fable had mostly the same pre-training data as Opus, and it's likely they're distilled from the same source. The difference is that it's a larger model with more post training on "dangerous" stuff they didn't want in the core model, and "long" task RL.