Grok 4.5 shipped, debuting at #4 on GDPval-AA at a tenth the cost of its superiors and seizing #1 on AutomationBench as the first model to complete over half of real SaaS workflow objectives without breaking business rules.
The credited mechanism is a tight loop: a smarter model behaves better, and better behavior unlocks more intelligence. One asterisk — the run accidentally ingested the Cursor codebase, benchmark tasks included.