Claude Fable 5 only achieved mid-tier results on coding benchmarks, and critics suggest data leakage may have inflated earlier scores.

This raises questions about the reliability of AI evaluation methods.

AI-generated: summaries written by AI from the linked sources. How we use AI