The Introducing Grok 4.7 post positions the model as SpaceXAI’s most powerful system for coding and knowledge work — “twice as fast, at half the price of comparable models” in their framing — while explicitly matching Grok 4.6 on list price and speed for the standard variant.
What SpaceXAI listed for Grok 4.7
Per the launch post, Grok 4.7 uses a new, larger base model than Grok 4.6, trained with a longer reinforcement-learning run on a harder mix of tasks weighted toward problems that take many hours. The model is described as better at verifying its own work and managing longer context, and as natively understanding the Grok Bot harness for conversational and general knowledge work.
Vendor benchmark table on that post (attribute to xAI/SpaceXAI only — not this desk’s independent run): CursorBench 4.0 at 46.3% versus Grok 4.6’s 40.4%; Terminal-Bench 4.0 at 38.0% versus 20.3%. Fable 5.1 Max remains ahead on several of those rows (CursorBench 51.8%, Terminal-Bench 57.9%). DeepSWE v1.1 is listed at 71.0%* for Grok 4.7 with a high-effort footnote.
“Grok 4.7 is our most capable model for coding and knowledge work. It works longer on difficult tasks, checks its own work more carefully, and comes with our best-calibrated safeguards to date. Served at the same price and speed as Grok 4.6, it is highly competitive in its class.”SpaceXAI / xAI — Introducing Grok 4.7
What it costs (from xAI docs)
These are USD rates from docs.x.ai/docs/models. Do not quote them as CAD. CAD is on the invoice. Long-context pricing applies when the prompt reaches the listed threshold — the higher rate then bills all tokens in the request.
| Line | Input / 1M | Output / 1M | Notes |
|---|---|---|---|
| grok-4.7 (<200k prompt) | $2.00 | $6.00 | 500k context · cached in $0.50 |
| grok-4.7 (≥200k prompt) | $4.00 | $12.00 | Same context · cached in $1.00 |
| grok-4.6 (compare) | $2 / $4 | $6 / $12 | Same banded rates as 4.7 |
| Fast variant | — | — | 2× output speed at 2× price (launch post) |
Docs also state the knowledge cut-off date of Grok 4.7 is May 2026. For current events, enable server-side search tools rather than assuming the base model knows today’s Alberta news.
What Alberta operators should do
For AI automation Alberta and AI consulting Alberta teams already in Cursor, the operator move is straightforward: switch the model picker to Grok 4.7 at the same $2/$6 band as 4.6 and re-run your longest repo tasks. Use Grok Build or the Grok API when the workflow leaves the IDE. Keep a human review gate on merges that touch billing, credentials, or customer data.
Foreign hosted API is still not a private Alberta path. If the codebase or prompts include client health notes, tender packages, or plant credentials, pair this briefing with Opcelerate’s AI consulting and private AI security lanes — and keep reading The Super Intelligence Times for source-backed model cards. Quote USD as published; expect CAD on the invoice.
The guardrail
Same price as 4.6 is not permission to skip evaluation. Vendor CursorBench and Terminal-Bench gains are SpaceXAI’s numbers; Fable 5.1 still leads several of those rows on their own table. Do not treat a coding-model upgrade as AGI or ASI. Budget the ≥200k band before you paste a whole monorepo, and keep regulated workloads off public paste endpoints.