BREAKING: Grok 4.5 takes the #1 spot on a benchmark that tests how well AI can predict whether a drug could cause liver damage, one of the biggest reasons medicines fail during development.