xAI
Grok 4.7 launches broadly, priced like Grok 4.6
xAI released Grok 4.7 on Sept. 21, 2026, in general availability at the same per-token price as Grok 4.6, alongside a faster tier that costs twice as much.
xAI released Grok 4.7 in general availability, calling it the company’s most capable model to date for coding, agentic tasks and knowledge work, while holding the per-token price at the same level as its predecessor, Grok 4.6.
The company said Sept. 21, 2026, that Grok 4.7 is available today in Cursor and Grok Build, and through the Grok API, third-party coding harnesses, model routers and cloud platforms. That is a broad rollout, not a preview or a limited release.
Grok 4.7 costs $2 per million input tokens and $6 per million output tokens, xAI said, the same rate it charged for Grok 4.6. xAI’s developer documentation at docs.x.ai showed matching figures, with no separate rate listed for longer context, when checked Sept. 22, 2026. A faster-serving variant of Grok 4.7 runs at twice the output speed for twice the price, according to xAI.
The model carries a 500,000-token context window and the model ID grok-4.7, xAI’s documentation showed.
xAI said Grok 4.7 “uses a new, larger base model compared to Grok 4.6” and was trained with “a longer reinforcement learning run on a harder mix of tasks,” weighted toward problems that take many hours to complete, according to the company.
On self-reported benchmarks, xAI said Grok 4.7 scored 46.3% on CursorBench 4.0, 38.0% on Terminal-Bench 4.0 and 56.7% on HealthBench Professional. The company has not published independently verified results for those tests.
xAI also said Grok 4.7 is “the strongest model they’ve tested on refusals and jailbreak resistance,” pointing to a HackerBench v0.3 result in which the model let through only 3.3% of risky dual-use prompts, by the company’s own account.
Analysis
xAI is selling Grok 4.7 on capability, not on price. Matching Grok 4.6’s rate while claiming a larger base model and a longer training run is a bet that the gains show up in what the model can do rather than what it costs to run, at least for developers who stay on the standard tier instead of the faster one, which charges double.
The benchmark and safety numbers come from xAI alone. CursorBench, Terminal-Bench, HealthBench Professional and HackerBench are tests xAI chose, ran and reported on the model it built. That does not make the scores wrong, but Grok 4.7’s standing next to other models on those same benchmarks is not yet something an outside party has confirmed.
Where Grok 4.7 sits in xAI’s model lineup going forward is also not settled by this release alone. A company calling its newest model the most capable it has shipped is a claim about today, not a promise about what comes next.