Grok & Cursor Watch: Grok 4.7 available on the xAI API

Series: Grok & Cursor Watch · 28 September 2026 (JST)

Summary

xAI API release notes published on 2026-09-27 say Grok 4.7 is now available as grok-4.7. The notes describe it as SpaceXAI's frontier model for coding, agentic tasks, and knowledge work; that description is a vendor claim and is not independently shown in the note. It has a 500k context window, takes text and image inputs, returns text only, and has no text output limit. Pricing is $2, $0.50, and $6 per 1M tokens for input, cached input, and output below 200k prompt tokens, and $4, $1, and $12 above that threshold. Reasoning effort supports low, medium, high (the default), and xhigh; on the Responses API, grok-4.7 always returns reasoning.encrypted_content even when include does not list it, and the notes also say it is served on a US regional endpoint.

Why it matters

The price schedule doubles once prompts pass 200k tokens, so long-context coding and agent calls cost more than shorter ones. Responses API clients should expect encrypted reasoning content on every grok-4.7 reply, not only when they request it. The note supplies no independent evaluation, so the frontier-model positioning remains unproven outside the vendor's wording.

Source

xAI API Release Notes

Published 2026-09-27.

Label: Grok & Cursor Watch

Publisher: @sangojpnet / X @sangojpn