The xAI API, powered by Grok AI models, now supports various file formats, including text, PDF, CSV, JSON, and more.
The xAI API also supports multi-turn conversations with files
Photo Credit: xAI
Elon Musk's xAI has now added file support in its Grok-powered application programming interface (API). With this, developers can now offer this capability in the artificial intelligence (AI) apps and tools they develop and release. While most major AI companies already offer this feature, xAI has now joined the ranks as well. The company says that the API will now support a wide range of file formats, and it will also allow users to attach multiple files and have conversations about them.
xAI has been rapidly expanding its features. Recently, the company upgraded the Grok Imagine feature to allow users to type detailed prompts and generate videos based on them. Both instruction following, complex prompt understanding, and the output fidelity has been improved. Now, the Musk-led company is offering file support to developers via its Files API.
On Tuesday, Musk reshared a post on X (formerly known as Twitter) highlighting that the new Files API will allow developers to offer the ability to upload and attach a document and answer queries based on it. To analyse files, the API uses an agentic tool calling function. This is why the feature is only available with Grok 4 and Grok 4 Fast AI models.
At present, the API supports a large number of file formats, including plain text, markdown, Python, JavaScript, CSV, JSON, PDF, and more. However, the maximum file size it accepts has been limited to 48MB. Also, batch requests are not possible while pulling a request from files.
Coming to the features, it supports multi-file support and multi-turn conversations. Additionally, the code execution tool has been integrated to enable the underlying AI model to perform advanced data analysis, statistical computations, and more. The model can also write and execute Python code that processes your files directly.
Notably, the Files API charges $10 (roughly Rs. 885) per 1,000 tool invocations. Each time the model searches the documents, it counts as one tool invocation. These charges are applied in addition to standard token costs. File-based chats use prompt tokens, reasoning tokens, completion tokens, and cache tokens (for repeated document content).
Get your daily dose of tech news, reviews, and insights, in under 80 characters on Gadgets 360 Turbo. Connect with fellow tech lovers on our Forum. Follow us on X, Facebook, WhatsApp, Threads and Google News for instant updates. Catch all the action on our YouTube channel.