Source: Models and pricing (xAI docs)

xAI's Grok model list with context windows, per-token prices, cached-input prices and the 200K long-context price step.

Contents
  • What: official API documentation page
  • Publisher: xAI
  • Date: accessed 2026-11-14
  • Original: Models and pricing

Summary

Lists Grok text models, their context windows and prices, and explains model aliasing.

Key claims

  • Grok 4.7 (recommended for code and chat): 500K context, $2/$6 per MTok below 200K prompt tokens ($0.50 cached input), doubling to $4/$12 at or above 200K; knowledge cutoff May 2026.
  • Grok 4.3 and Grok 4.20 variants have 1M context at $1.25/$2.50 ($0.20 cached input); grok-4.20 ships reasoning, non-reasoning and multi-agent variants. Grok 4.5 caches at $0.30, so cached input runs 0.15x to 0.25x of base.
  • The page lists no batch discount; it notes that not every model accepts Batch API requests.
  • grok-build-0.1: 256K context, $1/$2.
  • Once a prompt reaches 200K tokens, the higher rate applies to all tokens in the request.
  • Aliases: <model> is latest stable, <model>-latest is latest, <model>-<date> is pinned.
  • logprobs are not supported on grok-4.20 and newer.

Feeds

Frontier model landscape · Model selection

Suggest changes →