Grok 4.7 arrives with a bigger base model and no price rise

single source· 1 articles · confidence: low · first seen 2026-09-22 04:10 UTC

What this means for you

Nothing to do yet. The switch is free — same $2/$6 rate, same speed as 4.6 — but no scores are published, so there is no evidence the larger base model or longer training run changed anything you would notice. Test it against 4.6 on your own tasks before migrating.

SpaceXAI has released Grok 4.7, its new flagship model, at the same $2/$6 rate and the same serving speed as Grok 4.6. It is built on a larger base model and a longer reinforcement learning run (training against a reward signal rather than worked examples alone). It ships as a hosted model, callable as grok-4.7. The release note gives no benchmark scores, no evaluation date and no context limit, so what the larger base model bought is unmeasured.

Models in this story

Key facts

  • ·SpaceXAI released Grok 4.7 as a flagship model for coding, agentic tasks and knowledge work. source
  • ·Grok 4.7 keeps the same $2/$6 rate as Grok 4.6; the source does not state the units. source
  • ·It serves at the same speed as Grok 4.6. source
  • ·It is built on a larger base model and a longer reinforcement learning run than 4.6. source
  • ·It is available as a hosted model callable as grok-4.7. source
  • ·The release note carries no benchmark scores and no evaluation date. source

What the sources say

  • MarkTechPost — Short release note confirming the model's price, speed and hosted availability, with no evaluation results.

Sources

The original reporting. Follow these — they did the work.

← the wire