Microsoft and xAI Bring Grok 4 to Azure AI Foundry

Grok 4 features a 128,000-token context window, tool use, and integrated web search.

Topics

  • Microsoft Corp. said it has partnered with Elon Musk’s xAI to make Grok 4, its most advanced large language model, available on Azure AI Foundry.

    The release brings reasoning-focused AI to enterprises with governance, content safety, and deployment controls built in.

    Grok 4 features a 128,000-token context window, tool use, and integrated web search.

    Microsoft said deployments through Azure AI Foundry have Azure AI Content Safety enabled by default.

    The companies said they worked over the past month on responsible AI evaluations, emphasizing accountability as well as performance.

    Trained on xAI’s Colossus supercomputer, Grok 4 builds on reinforcement learning and multi-agent systems and offers a ‘think mode’ for step-by-step reasoning.

    Early tests suggest strong performance in math, science, humanities, and multi-step problem solving. The extended context window allows processing of long legal files, research papers, or codebases, subject to token limits.

    For industries that demand rigorous, step-by-step answers, such as research, education, or complex troubleshooting, this reasoning-first approach is a game-changer.

    One of Grok 4’s defining technical advances is its ability to process vast amounts of information without losing track. Its extended context window allows it to take in hundreds of pages of documents, long academic papers, or even entire codebases, and then analyze them holistically.

    This means enterprises no longer need to split up documents or manage fragmented prompts. A legal team could upload complete case files, a university could analyze entire research journals, and developers could parse legacy codebases in one shot. Importantly, Grok 4 doesn’t just store this information; it uses what xAI calls “smart memory” to prioritize the most relevant details.

    Alongside Grok 4, Azure AI Foundry also offers specialized variants to cater to different enterprise needs. Grok 4 Fast Reasoning is optimized for logic-heavy tasks that demand deeper analysis and complex decision-making.

    Grok 4 Fast Non-Reasoning is designed for quick, lightweight operations such as summarization or classification where speed matters more than detailed reasoning. Meanwhile, Grok Code Fast 1 is tailored specifically for developers, excelling in code generation and debugging across multiple programming languages.

    Together, these variants give enterprises the flexibility to choose the right model for their specific workflows.

    Elon Musk Introduces ‘Grokipedia’

    In a separate development, Elon Musk on Tuesday, 30 September, announced that xAI is working on Grokipedia, an open-source knowledge repository designed to rival Wikipedia.

    Musk described it as “vastly better than Wikipedia,” with unrestricted public access and no usage limits. More than just a reference tool, Grokipedia is envisioned as a cornerstone in xAI’s broader mission to understand the universe, positioning it as both a knowledge hub and a bold step in the company’s grander roadmap.

    Topics

    More Like This

    You must to post a comment.

    First time here? : Comment on articles and get access to many more articles.