Comment by parker-3461

Makes me wonder if Anthropic is really having issues with allocating compute (see recent deals with xAI and SpaceX). From available benchmarks, it seems like similar results should be possible with GPT 5.5 Pro or Opus 4.7 (with specific cybersecurity trained models).