Five surfaces, one model family
Grok is not one chat box. In 2026 you build across chat, web Build, CLI agent, API, and Imagine.
xAI ships Grok as a product stack. Flagship text/coding is Grok 4.5 (configurable reasoning, large context, strong agentic tool use). Media goes through Imagine. Agentic coding is Grok Build — in the browser (app builder with live preview + *.grok.me publish) and as a local CLI agent.
The five surfaces
- Chat (grok.com / apps) — think, research, draft, debug in conversation; real-time search; voice.
- Grok Build (web) — plain-language app builder: apps, games, dashboards, sites with live preview and publish.
- Grok Build CLI — terminal coding agent (TUI + headless). Plan → review → approve. Reads your repo.
- API (api.x.ai) — grok-4.5 for your products, IDEs, and custom agent loops; OpenAI-compatible paths exist.
- Imagine API / product — image and video generation & editing for assets and creative work.
Mental model
Chat explores. Build (web) ships demos and products without a local toolchain. CLI owns serious repo work with diffs and plan mode. API embeds Grok into your software. Imagine dresses the product. Pros mix all five in one week.
Exercise
Surface pick
Write three tasks from your life (e.g. “landing page”, “fix flaky test”, “hero image”). Assign each to one surface and one sentence why.
- 3 tasks listed
- Each mapped to chat / Build web / CLI / API / Imagine
- One sentence rationale each