xAI releases Grok 4 Fast

Background

Reasoning models can spend many tokens working through a problem, which raises the cost of each answer. A model that needs fewer of those tokens is one way to make reasoning cheaper to run.^[1]

What happened

On 19 September 2025 xAI released Grok 4 Fast, a cheaper reasoning model with a 2-million-token context window.^[1] One set of weights serves both reasoning and non-reasoning modes, which means users switch between them rather than between separate models.^[1]

Key details

xAI said the model uses about 40% fewer thinking tokens than Grok 4 for comparable results.^[1] Its API prices started at $0.20 per million input tokens and $0.50 per million output tokens.^[1]

On benchmarks, xAI reported 85.7% on GPQA Diamond, a set of hard graduate-level science questions, and 92.0% on AIME 2025, a set of maths competition problems.^[1] These are the company's figures, reported at launch.

The model was made available to all Grok users, including the free tier, so people without a paid subscription could use it in the Grok apps.^[1] A cheaper model that reaches the free tier means the model's reach depends less on how much a user pays.

A context window is the amount of text a model can keep in view, measured in tokens. A larger window matters most for long documents and for multi-step agent work. Benchmarks such as GPQA Diamond and AIME are sets of hard questions in science and mathematics. A model's score on them shows its accuracy on that type of question, not its performance on every task.

Impact

The release marked a shift in xAI's line toward a model built for efficiency as well as capability. For readers, the key facts are simple: a larger context window, fewer thinking tokens for similar results, a low starting price, and availability to free users. Each of these can be checked against the published documentation.