xAI: Grok 4.7Active

Create a ZenMux API key, set your OpenAI-compatible base URL to https://zenmux.ai/api/v1, and use x-ai/grok-4.7 as the model ID. See the ZenMux API quickstart for request examples and supported API protocols.

Grok 4.7 is suited to coding, code review, debugging, multi-step agentic workflows, and knowledge work that requires sustained reasoning. For production use, define the task, provide the relevant context, and evaluate its answers against your own codebase or documents; benchmark scores cannot predict results on every workflow.

Grok 4.7 accepts text and image inputs and returns text, so you can ask it to explain a screenshot or inspect a visual reference. To send an image with an API request, follow ZenMux’s multimodal input guide. For image creation or editing on ZenMux, use the separate Grok Imagine Image 2.0 model.

ZenMux applies different billing tiers based on prompt length. The current model page lists $2–$4 per million input tokens, $0.50–$1 per million cached input tokens, and $6–$12 per million output tokens. Check the displayed rates when planning large requests.

No. The 500K-token context window is a capacity limit, not a target prompt size. Send the files, documents, and conversation history relevant to the current task. Keeping context focused can make the result easier to check and helps control costs, particularly once a request reaches the higher-context pricing tier.

Grok 4.7 supports reasoning controls on ZenMux. Choose the configuration supported by your request protocol, then test the effect on your workload. ZenMux’s reasoning models guide explains how to pass reasoning parameters through its API.

Yes. Grok 4.7 supports function calling. Your application supplies the tool definitions, executes a requested function, and sends its result back to the model; the model does not gain access to your backend or codebase merely because it supports tools. The ZenMux tool calling guide explains this request flow.

ZenMux documents implicit prompt caching for the Grok model series, so requests do not need a cache_control parameter. Cache hits depend on reusable prompt content and are not guaranteed. Put stable instructions before changing content, and use ZenMux request logsto inspect token and cache-read details.

For later news or changing information, provide current source material or connect an appropriate retrieval tool, then check the answer against that evidence. A standard model request should not be treated as an automatic live search.