xai / grok-4-1-fast

## Overview Grok 4.1 Fast is a text-first chat model made by xAI for agentic work. It handles very long inputs, up to a 2,000,000-token context window. It can plan multi-step work, call tools, and then write a final answer. It also accepts images for tasks like OCR, screenshot understanding, and visual Q&A, but it returns text. ## What you can build - Customer support agents that read long policy docs and answer with consistent tone - Deep research assistants that browse the web and return cited summaries - Finance and ops copilots that follow multi-step workflows and tool outputs - Contract and report Q&A that stays inside a single long conversation history - Screenshot and diagram explainers for troubleshooting and onboarding - Structured data extraction pipelines that need predictable JSON output ## Inputs - Your request as plain text, written as an instruction or question. - Optional system-level instructions that set role, tone, and rules for the assistant. - Optional image attachments in JPG, JPEG, or PNG format for visual understanding. - Optional selection between an instant-response variant and a deeper-reasoning variant. - Optional web search mode so the assistant can fetch up-to-date information and cite sources. - Optional conversation identifiers so the app can keep chat history across multiple turns. - Optional maximum response length limit, measured in tokens, to cap how much text the model generates. ## Outputs - A text response that answers the prompt, often with step-by-step explanations when useful. - When web search is enabled, the response can include citations and a list of visited sources. - When a reasoning-oriented variant is selected, the model may spend extra tokens planning before answering, while still returning a normal text reply. ## Limitations - This model returns text. It does not generate images or video. - Long-context does not guarantee perfect recall. Important details can still be missed. - Low-quality image inputs can break understanding. Blurry photos, heavy compression, tiny text, or skewed scans can cause OCR errors. - If you do not enable web search, answers about recent events may be outdated. - The model decides how it reasons. There is no separate “reasoning effort” dial for this model family. ## Safety & compliance - You must follow xAI’s Acceptable Use Policy when you prompt the model and when you use outputs. - The model is designed to refuse requests that involve illegal or harmful activity, including sexual exploitation of minors, weapons development, and self-harm content. - Do not try to bypass safety controls or prompt the model to produce disallowed content. - You are responsible for ensuring you have rights to any text or images you submit.