Is Opus 4.5 really ‘the best model in the world for coding’? It just failed half my tests

Here’s what happened when I pushed Anthropic’s new model through some simple development tasks.