Safety

AI-generated text

Sonnet 5 approached Opus 4.8 production safety level with fine-tuning

The Sonnet 5 model was fine-tuned on an early Opus 4.8 checkpoint and reached safety metrics close to the production Opus 4.8; this initial test suggests a model may be able to align stronger…

Sonnet 5 approached Opus 4.8 production safety level with fine-tuning

The Sonnet 5 model was fine-tuned on an early Opus 4.8 checkpoint and reached safety metrics close to the production Opus 4.8; this initial test suggests a model may be able to align stronger successors, which could affect alignment strategies and scalability.