Grok 4.6's cache discount is only 75%, pricier than rivals for non-coding use

brandon_galang · x · 2026-08-22

A user notes Grok 4.6 is deceptively expensive for non-coding workloads like general research: xAI bills cached input at $0.50/1M tokens vs $2 for non-cached — only a 75% discount, while most other models offer 90%. Worth noting if you use Grok 4.6 for cost savings.

Original post →

More from Models

Models channel →