SpaceXAI: Grok 4.7

SpaceXAI: Grok 4.7 is a text generation model from xAI, listed on api14 under the model id x-ai/grok-4.7. It sits in the text category and is described by its provider as the flagship model for coding, agentic tasks, and knowledge work, succeeding Grok 4.6. The provider release date is 2026-09-21.

The model accepts text, image, and file inputs, and it returns text. Its context window is 500000 tokens, and the maximum output is 450000 tokens, so very large inputs and long generations are both possible within a single request. Supported parameters include include_reasoning, reasoning, and reasoning_effort, which relate to the model's reasoning behaviour, alongside logprobs and top_logprobs for token level inspection.

Those capabilities point to a few practical uses. The provider highlights long-running software engineering tasks, verifying its own work, and agentic workflows, so the model fits code generation and multi-step tool use where a task runs over many turns. Because it accepts images and files, it can also be given documents or screenshots alongside a prompt, which suits knowledge work such as summarising or extracting information from mixed material. The large context window helps when the whole codebase or document set must be in view at once.

Through api14 you call it with the OpenAI-compatible chat completions endpoint and the model id x-ai/grok-4.7. The request supports max_tokens, temperature, top_p, seed, response_format, structured_outputs, and the reasoning controls (reasoning, reasoning_effort, include_reasoning). It also supports tool calling via tools and tool_choice, and token probability output via logprobs and top_logprobs. Set the parameters you need in the request body as you would with any OpenAI-compatible client.

One thing to weigh is the output ceiling: 450000 tokens is generous but not unlimited, so extremely long generations may need to be split or continued across requests. If a task does not involve images or files, or does not need the reasoning and tool features, a simpler text model may be sufficient. For coding, agentic work, and large-document tasks, this model's combination of a 500000 token context and tool support is the reason to choose it.