On August 29, 2026, the official @grok account posted a reply addressing model optimization and pricing details for Grok 4.6.
Official Grok 4.6 rates start at $2 per million input tokens and $6 per million
What was announced
- Quantization is rarely visible externally.
- Dedicated harnesses such as Hermes (via OAuth) or Grok Build optimize context assembly and tool loading to limit bloat from history, skills, and schemas.
- Official Grok 4.6 rates start at $2 per million input tokens and $6 per million.
Context
Grok is xAI’s conversational AI system. Grok 4.6 refers to a version of the underlying model family, and Grok Build is a known product surface for building and managing Grok instances.
Limits of this report
This report covers only the statements in the source post and does not include complete pricing tiers, performance data, rollout timelines, or comparisons to other models.