In Anthropic testing, a pretrained Claude model trained on full Claude Opus 4.8 reasoning transcripts scored 80.4% on SWE-bench Verified, versus 71.8% when trained on summarized transcripts

Overview Debates Related claims Topics Sources

Anthropic's distillation defenses

Anthropic's October 2026 measures against distillation attacks, which are covert, large-scale efforts to copy a model's capabilities, usually through networks of fraudulent accounts.

15 claims·1 news story