xAI review

Prices are per token, with input fees of $1 to $4 and output fees of $2 to $12 across the catalogue. Neither figure is marked as an introductory rate, so budget with those numbers from the first call. Replies carry the larger share, so long generations cost more than the input rate suggests.

The largest context window listed is 1,000,000. That figure is the ceiling for the catalogue, so a specific Grok release may sit lower and it is worth checking the model you plan to call. At that size, long documents and large codebases fit in one pass.

Yes. The API is OpenAI-compatible, so base URLs, SDK settings and request code move over with a new endpoint and key. Teams on an existing OpenAI-shaped stack can add Grok without rewriting handlers, and the catalogue behind grok 4.7, grok 4.6 and grok 4.5 is reachable through the same interface.

The list behind grok 4.7, grok 4.6 and grok 4.5 spans several chat and reasoning releases, a multi-agent variant, a build-oriented model, and image, video and voice models. Text and media therefore come from one provider, and the fast-thinking voice model sits alongside the newest reasoning entries.

It sits at the expensive end of the field. Input fees of $1 to $4 and output fees of $2 to $12 run well above rivals, while Salad lists input fees of $0.06. When token volume is the main cost line, a cheaper route cuts the bill noticeably.

It suits developers who want a 1,000,000 context window, want Grok specifically and already speak an OpenAI-compatible API. It suits them less when volume dominates, because $1 to $4 and $2 to $12 set the ceiling on what a busy month costs.

A gateway sells breadth, a model vendor sells depth on its own family. Supascale AI lists DeepSeek V4 Pro, GPT 5.6 Sol and GPT 5.6 Terra across many vendors, while xAI stays within grok 4.7, grok 4.6 and grok 4.5. Pick the gateway when model optionality matters, and xAI when a 1,000,000 window on a single family matters more.