grok-4.7 becomes flagship and a 2x Grok 4.7 Fast card lands outside the public API
grok-4.7 replaces grok-4.6 at an unchanged $2.00 in / $6.00 out per 1M tokens, but a new Grok 4.7 Fast card bills 2x — Cursor and Grok Build only, never the public xAI API.
Text API flagship grok-4.6 at $2.00 input / $0.50 cached / $6.00 output per 1M tokens (500k context); long context $4.00 / $1.00 / $12.00. US regional endpoint at 1.1x covers grok-4.6 only. No fast-infrastructure variant on the rate card.
Text API flagship grok-4.7 at the same $2.00 / $0.50 / $6.00 short context and $4.00 / $1.00 / $12.00 long context (500k context, knowledge cutoff moved to May 2026); grok-4.6 stays published at identical rates. New Grok 4.7 Fast section at 2x standard rates: $4.00 input / $1.00 cached / $12.00 output per 1M below 200k prompt tokens, $6.00 / $1.50 / $18.00 above. US regional endpoint now covers grok-4.7 and grok-4.6, printed for grok-4.7 as $2.20 / $0.55 / $6.60 short context and $4.40 / $1.10 / $13.20 long.
Proof of change
xAI's pricing pages, as we captured them on two dates.
These values appear in only one of the two captures. That can mean a page-layout difference rather than a price move — read the images, not just the list.
Showing the whole page as captured — scroll either panel, or open it at full size.
Both images are our own captures, taken on the dates shown. The pixels are unmodified — nothing is retouched, and where a region is highlighted the marker is drawn over the image, not into it. Open either capture to see it at full resolution.
xAI turned its flagship over twice in a month. grok-4.7 took the top row of the Text API rate card on 21 September 2026, one day after grok-4.6 landed there — and at exactly the same headline: $2.00 per 1M input, $0.50 cached, $6.00 output, on a 500k-token context with the long-context band starting at 200k. grok-4.6 is not retired; it drops to a second row at identical rates, so the published ladder now carries three flagships (4.7, 4.6, 4.5) sharing one $2.00 / $6.00 price point. For a buyer the model upgrade is free — the knowledge cutoff moves from February 2026 to May 2026 and nothing on the invoice changes.
The genuinely new price is a section that did not exist before: Grok 4.7 Fast pricing (Cursor and Grok Build only). It is described as “the same Grok 4.7 model served on faster infrastructure, at twice the standard token rates,” and it is gated — “not available on the public xAI API,” and excluded from Grok Build’s free tier. Below 200k prompt tokens it bills $4.00 input / $1.00 cached input / $12.00 output per 1M; above 200k it bills $6.00 / $1.50 / $18.00. The long-context step is only 1.5x the standard long-context column rather than 2x, because standard grok-4.7 long context is already $4.00 / $12.00. Cursor bills its own fast variant through the Cursor plan, so the same model can reach a team on two different invoices at two different rates.
Latency is now a priced axis on three fronts at once: Priority Processing (2x, public API), Grok 4.7 Fast (2x, partner channels only), and the US regional endpoint (1.1x for data residency, which this release widened from grok-4.6 alone to grok-4.7 and grok-4.6). Everything else on the card holds — agentic tools including the 21 September shift of X Search to $5 per 1k posts plus $10 per 1k profiles, Imagine, Voice, the 20% Batch discount, storage and downloads. The consumer Grok app is untouched and still runs Grok 4.6 at $0 / $30 / $100 per month.
One day after grok-4.6 became the flagship, grok-4.7 replaced it on the Text API rate card at an identical $2.00 input / $0.50 cached / $6.00 output per 1M tokens short context and $4.00 / $1.00 / $12.00 long context (500k context, knowledge cutoff moved from February 2026 to May 2026). grok-4.6 does not go away — it drops to a second row at exactly the same rates, so the published ladder now carries three flagships (4.7, 4.6, 4.5) at one $2.00 / $6.00 headline. The genuinely new price is a separate section, "Grok 4.7 Fast pricing (Cursor and Grok Build only)": the same model on faster infrastructure at twice the standard token rates — $4.00 in / $1.00 cached / $12.00 out per 1M below 200k prompt tokens, $6.00 / $1.50 / $18.00 above — explicitly not available on the public xAI API and excluded from Grok Build's free tier, with Cursor billing its own fast variant through the Cursor plan. The US regional endpoint widens from grok-4.6 only to grok-4.7 and grok-4.6, and now prints its 1.1x result outright for grok-4.7: $2.20 / $0.55 / $6.60 short context, $4.40 / $1.10 / $13.20 long. Everything else holds — agentic tools, the 2026-09-21 X Search result-metering switch, Imagine, Voice, Batch (-20%), Priority (2x), storage and every consumer plan are unchanged, and the consumer Grok app still runs Grok 4.6. (Source: docs.x.ai/docs/pricing and docs.x.ai/docs/models, both last updated September 21, 2026; live capture 2026-09-22.)