xAI Ships Grok 4.7 to Its Public API at $2/$6 per Million Tokens

xAI has shipped Grok 4.7 to its public API, pricing the model at $2 per million input tokens and $6 per million output tokens with a 500,000-token context window — a release that positions the company on price and context rather than on the benchmark chase its rivals have been running.

The update arrived on September 21 and came with a quieter but telling change: the documentation now sits behind new "SpaceXAI Docs" branding at docs.x.ai, a nod to the consolidation of Elon Musk's AI and space ventures and a reminder that xAI's distribution advantages are not limited to the chatbot market.

The friction is on the integration side. Pricing, context limits and reasoning controls have all moved enough since Grok 4.5 and 4.6 that existing integrations built on the older listings can break, and developers have taken to flagging the discontinuities. For a company competing against rivals with far larger enterprise footprints, changing terms underneath existing workloads is a cost that shows up as developer trust rather than as a line on a pricing page.

The strategic picture is worth separating from the release itself. In a week dominated by OpenAI's GPT-6 Sol and Luna and Anthropic's Claude Opus 5.5 — both framed explicitly around cutting cost per token, with Anthropic cutting the price of its flagship by roughly 40% — xAI's move is less a counterpunch than a bid to stay in the comparison set. A $2/$6 price and a 500,000-token context are competitive enough to keep xAI on procurement shortlists, even if the company is not winning the capability headlines.

xAI has also found itself in an unusual position on the policy side. A new antitrust suit filed in the Northern District of California names SpaceXAI alongside OpenAI, Anthropic and Google, accusing the four of an illegal agreement to slow AI development after their leaders publicly endorsed a call for cooperation on frontier safety. xAI is thus being sued for promising to go slower while shipping models on a faster cadence than most competitors — an inconsistency that opponents of the "pacing" framing will happily point to.

For developers, the practical takeaway is narrower. Grok 4.7 is now the current listing, it is priced competitively against the mid-tier of the market, and anything built on the previous generations should be re-tested rather than assumed forward-compatible. In a market where model swapping has become the norm, that is close to business as usual — which is itself the point: the frontier labs are now competing on whose contract terms change least painfully, not only on whose model is best.