Grok 4.7 has shipped
Confirmed vs rumor
As of 2026-10-07, Grok 4.7 is out: xAI (SpaceXAI) released it on 2026-09-21, and the API model grok-4.7 went live the same day. The official model pages and Release Notes list 500k context, a May 2026 knowledge cutoff, $2 input / $0.50 cached input / $6 output per 1M tokens below 200k prompt tokens ($4 / $1 / $12 above), and reasoning effort low / medium / high (default) / xhigh. The launch post carries self-reported benchmarks, which this page does not transcribe; the timeline below records the delays before the listing. QCode does not offer Grok.
Updated 2026-10-07
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
What we can say today
Official launch post
xAI released it on 2026-09-21; as of 2026-10-07 docs.x.ai has a standalone grok-4.7 model page (500k context, starting at $2/$6).
Founder quote only
Musk (2026-07-28): a 2.1T model a few weeks after 4.6; better except slightly slower, more token-efficient.
The prior xAI tier
Grok 4.6 scored 61 on AA Intelligence, matching GPT-5.6 Sol, at $2/$6.
Rumor window closed
xAI released Grok 4.7 on 2026-09-21, so the rumor window is closed; the pre-listing delays stay in the timeline below.
What people mean by Grok 4.7
xAI's catalog calls Grok 4.7 its flagship (“Our flagship model for code and everything else”). The launch post says 4.7 uses “a new, larger base model compared to Grok 4.6”, but as of 2026-10-07 neither the model page nor the announcement gives a parameter count, so Musk's 2.1T remains a quote. The official price is “starting at” $2 / $6.
Why this query is hot
In late July 2026 the 4.6 and 4.7 teasers landed together, which pushed both names into search at the same time. With 4.6 shipped on 2026-08-12, the open questions became when 4.7 arrives and whether it can catch Fable 5. The first one xAI answered with its 2026-09-21 launch post; the second currently exists only as the vendor's own table, which this page does not transcribe.
Public timeline
Musk, 2026-07-24: Grok 4.6 in ~2 weeks, Grok 4.7 in ~4 weeks.
Musk, 2026-07-28: 4.7 is the 2.1T model; better than 4.6 except slightly slower to serve, better token efficiency.
Grok 4.6 GA (later than the ~Aug 7 hint). Still no 4.7 card as of 2026-09-18 (the same capture does list grok-4-6 and the 4.20 family, so it is not a scraping miss). xAI's announcement for Grok 4.7 came out on 2026-09-21; since 2026-09-22 the official model catalog lists grok-4.7 (a standalone card was still absent then, and the announcement's self-reported scores are not transcribed here). Rechecked 2026-10-07: docs.x.ai now has a standalone grok-4.7 model page that states a May 2026 knowledge cutoff.
Confirmed vs rumor
Confirmed
Confirmed word for word on official pages: Grok 4.7 released 2026-09-21 as API model grok-4.7; 500k context, text and image input with text-only output; $2 / $0.50 / $6 per 1M tokens (input / cached input / output) below 200k prompt tokens and $4 / $1 / $12 above; knowledge cutoff May 2026. Grok 4.7 Fast is the same model at 2x the standard rates (1.5x for long context), offered only in Cursor and Grok Build, not on the public xAI API. The founder's 2026-07-28 words — 2.1T / slightly slower / more token-efficient — are a description, not a spec sheet.
Rumour · specs listed, scores self-reported
The date is official now: the announcement is stamped Sep 21, 2026. As for scores, the launch post has its own table — this page and the comparison page transcribe neither. The 2.1T figure is still the founder's 2026-07-28 words, not a spec sheet, and “3–4 more weeks plus SpaceX data” is still community talk. Context and price are listed, so they are no longer rumour.
What you can compare today
Grok 4.6 vs Fable 5 (real numbers)
Official 4.6 table: AA 61 vs 62, DeepSWE 65.9% vs 70%, CursorBench 69.9% vs 70.5%. Close, not a win.
Grok 4.7 vs Fable 5 (self-reported table only)
The launch post carries a self-reported benchmark table; neither this page nor the comparison page transcribes scores — only published specs and prices.
How to watch the launch
Watch x.ai/news rather than screenshot threads: the 2026-09-21 launch post is the source, and it puts the starting price at $2 / $6 (“priced starting at”). QCode does not offer the Grok line; to drive a coding agent on one key today, pick claude-opus-5-5, claude-sonnet-5 or gpt-6-sol from /models.
On QCode
Grok 4.7 is callable on xAI's own API (checked 2026-09-22). Our price catalogue lists seven Grok ids (4.3 / 4.5 / 4.6 / the 4.20 family and grok-build-0.1), all with zero billed calls in the last 30 days, and we make no availability promise for the Grok line. Frontier options currently on /models include gpt-6-astra, claude-opus-5, gpt-5.6-sol, deepseek-v4-pro and kimi-k3.
Grok 4.7 FAQ
Is Grok 4.7 released?
Yes. xAI released Grok 4.7 on 2026-09-21, “priced starting at” $2 / $6 per million tokens, as API model grok-4.7. As of 2026-10-07 QCode does not offer Grok; the launch post carries self-reported benchmarks, which this page does not transcribe.
When will Grok 4.7 come out?
It is out: the announcement is dated Sep 21, 2026. Musk's 2026-09-02 “10 days” and the earlier “few weeks after 4.6” both landed that day. For dates, x.ai/news wins.
Is Grok 4.7 2.1 trillion parameters?
Not officially. 2.1T comes from a founder post; as of 2026-10-07 neither xAI's model page nor the launch post states a parameter count. The announcement only says 4.7 uses “a new, larger base model compared to Grok 4.6”.
Is Grok 4.7 better than Claude Fable 5?
The 4.7 launch post carries a self-reported benchmark table (Fable 5.1 is one of its columns); this page does not transcribe scores and draws no “who is stronger” conclusion.
Should I wait for 4.7 or start with 4.6?
No need to wait: xAI released Grok 4.7 on 2026-09-21, starting at $2/$6. QCode does not offer the Grok line; for frontier calls on one key today, see the /models list.
Will QCode add Grok 4.7 on day one?
Not offered. As of 2026-09-29 QCode carries no Grok line — zero billed calls across it in the last 30 days. Once a vendor API passes reliability checks it can appear on /models; no promise before that.
Sources
Sources: the Grok 4.6 release note on x.ai/news (2026-08-12); founder posts as relayed by contemporary reporting (2026-07-24 and 2026-07-28); Artificial Analysis' Grok 4.6 article (2026-08-12). The 2026-09-18 fetch (docs.x.ai/docs/models, archived 2026-09-18) held no model card, pricing or API id for Grok 4.7; refetched 2026-09-22 the same catalog lists grok-4.7 as the flagship, with Context 500k tokens and $2.00 / 1M tokens, $6.00 / 1M tokens. The 2026-09-21 announcement “Introducing Grok 4.7” states “The model is priced starting at $2 per million input tokens and $6 per million output tokens.”; it also carries a self-reported benchmark table, none of whose scores are transcribed here. A standalone 4.7 model card was still absent in the 2026-09-22 pull. Refetched 2026-10-07: the “Introducing Grok 4.7” announcement on x.ai/news, the docs.x.ai Models catalog, the standalone grok-4.7 model page, the Grok 4.7 overview page and the Release Notes (the September 21 entry gives the below/above-200k prices and says Grok 4.7 Fast is offered only in Cursor and Grok Build).
Use today’s frontier mix
One key covers what is actually callable today — claude-sonnet-5, claude-opus-5-5 and gpt-6-sol on /models. xAI released Grok 4.7 on 2026-09-21; QCode does not offer the Grok line.
Related
How to buy Claude API credits
Console prepaid credits, auto-reload, payment methods, one-year expiry and Max/Team monthly credits, compared with QCode top-ups.
Grok 4.7 vs Fable 5
What is known and unknown about 4.7, side by side.
2026 AI Model Radar
Live vs rumored models on one board.
QCode is not affiliated with SpaceXAI, xAI, or Anthropic. Grok 4.7 specifications come from xAI's official model catalog; the launch post carries self-reported benchmarks, which this page does not transcribe.