Grok 4.7 uses ~3x the tokens per task as Astra, tests show

User tests indicate Grok 4.7 consumes nearly 3x the tokens per task as Astra and over 2x more than Grok 4.6, leading the poster to argue OpenAI is the only lab that seriously cares about token efficiency.

2026-09-22 ~ 2026-09-22 · 2 related posts