xAI's Grok 4.5 / Grok Build Undercuts Opus 4.8 by ~60%: $2.49 vs $11.80 per Task, but Hallucinations Are Up

TechTimes·high signal

Launched July 8 and trained on Cursor developer-session data, Grok 4.5 posts 83.3% on Terminal-Bench 2.1 and 64.7% on SWE-Bench Pro at $2/$6 per 1M tokens — roughly $2.49 per coding task vs $11.80 for Claude Code, using ~15,954 output tokens against Opus 4.8's ~67,020 on the same task. The catch, flagged independently, is higher hallucination rates, so the real decision is token-efficiency-and-price versus reliability, not raw benchmark position.