“Claude Opus 4.8 is more reliable and sharper in its judgment when performing agentic tasks.”
Anthropic says the new model is sharper, more honest, and four times less likely to wave through broken code. Every launch comes with the same chart claiming this one is finally the good one. The price did not move and the benchmarks come from the company selling the product. Better is what they always say. We will see what survives contact with real codebases.