AI Insights

Analysis of AI tools, LLMs, and the evolving machine learning landscape, written by practitioners.

articles

Claude Opus 4.8: Benchmarks, Alignment and What Actually Changed

Anthropic released Claude Opus 4.8 today at the same price as 4.7. It leads on SWE-Bench Pro (69.2%), Humanity Last Exam reasoning, computer use, and legal benchmarks. Misaligned behavior dropped to Mythos Preview levels. Here is every number.

Oleh KemMay 28, 2026