Back to News

Software Development · Sep 22, 2026

Grok 4.7 and the new wave of coding models: what it means for custom software

On September 21, xAI made Grok 4.7 available in its API. The company describes it as its frontier model for coding, agentic tasks, and knowledge work. It has a 500,000 token context window, accepts text and images, and supports several reasoning effort levels.

It arrived in the middle of a crowded month. Anthropic shipped Claude Fable 5.1 on September 1 and Claude Opus 5.5 on September 22. OpenAI shipped GPT-6 Astra on September 3. Every one of them led with coding.

When every major lab is competing on the same skill, that skill gets better and cheaper fast. For businesses, that skill is building software.

Why coding models matter to non-software companies

Most small and mid-size businesses are not software companies. But almost all of them run on a patchwork of software: a CRM that does not quite fit, spreadsheets doing the work of a database, a scheduling tool that does not talk to invoicing.

For years the answer to "can we get a tool built that fits how we work" was usually "not at your budget." Custom software was for companies with an IT department.

Strong coding models change that equation. A competent developer working with these tools can build, test, and ship small internal applications much faster than before. That opens up custom software for problems that were too small to justify it in the past.

What a long context window actually buys you

The 500,000 token context window is a technical detail with a practical meaning. It means the model can hold a large amount of code, documentation, or data in view at once.

For custom software, that matters when the model needs to understand an existing system before changing it. Most real business software is not built from scratch. It is connected to something that already exists. A model that can read more of that existing system makes fewer wrong assumptions.

The part the benchmarks do not cover

AI coding tools are very good at writing code. They are not responsible for deciding what to build, how it should handle edge cases in your business, or how it will be maintained in two years.

I have seen fast AI-assisted builds produce software that works on day one and becomes a problem on day ninety, because nobody thought about who owns it, how it gets updated, or what happens when the model provider changes its API. Speed of writing code is now cheap. Judgment about what to build and how to keep it running is still the scarce part.

Questions to ask before you build

  1. What process does this replace, and who uses it today? Talk to them first.
  2. What systems does it need to connect to? Integration is usually more work than the app itself.
  3. Who maintains it? Someone has to own updates, bugs, and security.
  4. What happens if the AI model it depends on changes? Plan for model updates the same way you plan for any vendor change.

What this means for your business

  • Custom software for small, specific problems is more affordable to build than it was a year ago.
  • The cost has shifted from writing code to designing the right solution and maintaining it.
  • Long context models are better at working with your existing systems, which is where most business software lives.
  • Choose builders who talk about maintenance and ownership, not just speed.

If you have a process held together by spreadsheets and workarounds, start by checking whether the tools you already have can do the job.

Sources