xAI Launches Grok 4.7 with NVIDIA Accelerated Computing Support for Coding and Knowledge Work

xAI launched Grok 4.7 on 21 September 2026, describing it as the company’s most capable model to date for coding and knowledge-work applications. The 500,000-token context window is the headline technical figure — large enough to process the equivalent of several full-length novels in a single prompt, which matters considerably for developers working across large codebases.

NVIDIA publicly congratulated xAI on the launch and stated that it supports the team through NVIDIA accelerated computing. The specific GPU hardware, data-centre locations, and scale of that deployment were not disclosed. But the public endorsement from one of the world’s dominant AI-chip suppliers does confirm that NVIDIA infrastructure underpins the model.

What xAI Is Claiming

xAI says Grok 4.7 can work for longer on difficult tasks, check its own work more carefully, and comes with what the company describes as its best-calibrated safety measures to date. Those are the company’s own characterisations. No independent benchmark ranking or third-party evaluation was available at the time of writing, so they should be treated as launch claims rather than established fact.

The model accepts text and image inputs and produces text-only output. It’s aimed squarely at software developers, people running automated or agentic tasks, and knowledge workers who need a model that can sustain effort across complex, multi-step problems.

Where You Can Access It

Grok 4.7 is available through the Grok API, the Cursor coding environment, Grok Build, third-party coding harnesses, model routers, and cloud platforms, according to xAI’s developer documentation. A faster variant — Grok 4.7 Fast — is described as delivering roughly twice the output speed at twice the price, though availability of that version is currently restricted to Cursor and Grok Build.

Pricing on the standard model is tiered by context length. Below 200,000 prompt tokens, xAI states the API costs $2 per million input tokens (around £1.58), $0.50 per million cached input tokens (around £0.40), and $6 per million output tokens (around £4.74). Above 200,000 tokens, those figures double: $4, $1, and $12 per million tokens respectively (around £3.16, £0.79, and £9.48). All figures are from xAI’s developer documentation and are stated in US dollars; UK-pound pricing, VAT treatment, and confirmed availability in the United Kingdom have not been independently verified. XAI says the base price and speed are the same as the previous Grok 4.6 model, though the exact release date of Grok 4.6 was not independently confirmed in the available material.

NVIDIA’s Role

NVIDIA’s public statement of support is notable less for what it reveals than for what it confirms. The company said it supports the xAI team with NVIDIA accelerated computing — a clear signal that the model runs on NVIDIA hardware, even if the specifics of the contract, the GPU generation involved, and the scale of deployment remain undisclosed.

Jensen Huang, NVIDIA’s chief executive, has previously spoken about the company’s ambitions in AI infrastructure. NVIDIA’s involvement across multiple frontier AI model launches this year reflects how central its hardware has become to the industry’s output.

What Competitors and Independent Reviewers Would Need to Establish

xAI’s claim that Grok 4.7 is “the most capable” model for coding and knowledge work is the company’s own description. It hasn’t been independently tested against comparable models — OpenAI’s GPT-4o, Anthropic’s Claude Opus 4, or Google’s Gemini 2.5 Pro, for instance. Benchmark results, where they exist, are often produced or commissioned by the companies themselves, which limits their usefulness as neutral comparisons.

Any organisation thinking about deploying Grok 4.7 — or any generative AI model — would need to assess reliability, data privacy arrangements, copyright exposure, cybersecurity risk, potential bias, and regulatory compliance before putting it to work on sensitive tasks. That’s not a reason to avoid it, but it is a reason to look carefully before committing.

What This Means for Kent Residents

Kent residents and businesses can access Grok 4.7 through the Grok API and platforms such as Cursor, subject to availability and the applicable terms of service, though UK pricing and data-hosting arrangements have not been independently confirmed. Software developers, technology businesses, and students at Kent’s universities may find it a useful addition to the generative-AI coding tools already on the market. Anyone considering using the service — whether for personal projects or business purposes — should check how their data is processed and stored before entering confidential or commercially sensitive information into the platform.