SpaceXAI: Grok STT 1.0

SpaceXAI: Grok STT 1.0 is a speech-to-text model from xAI. It sits in the transcription category, and its job is narrow and clear: take audio in and return a transcription out. It is not a general chat or reasoning model, so you should treat it as a dedicated component in a pipeline rather than something you hold a conversation with. The provider release date is 2026-07-23.

The model accepts audio as its input modality and produces a transcription as its output. According to the provider description, it supports transcription with word-level timestamps, optional speaker diarization, and multichannel audio. Word-level timestamps let you align text back to specific points in the recording, which matters when you need to sync subtitles, jump to a moment in a call, or highlight a phrase in a player. Speaker diarization is optional, so you can turn it on when you need to know who said what and leave it off when a single flat transcript is enough. Multichannel audio support means recordings with more than one channel can be handled rather than downmixed blindly. No context length is listed for this model, and the details block reports a max output of 0, so there is no meaningful token window figure to plan around here. The supported parameters list is empty.

The practical uses follow directly from those capabilities. Meeting and call transcription is the obvious one, and with diarization enabled you get a transcript that distinguishes participants, which is what you want for notes, summaries, or compliance records. Word-level timestamps suit subtitle and caption generation, searchable media archives, and any interface where a user clicks a line of text and expects playback to seek to that exact moment. Multichannel audio is useful for recordings captured from separate sources, such as a room mic and a line feed, or for contact center setups where each side of the conversation arrives on its own channel. Because the output is a transcription rather than generated prose, it also fits cleanly into downstream steps: you transcribe first, then send the text elsewhere for summarization, translation, or indexing.

You call it through api14's OpenAI-compatible API using the model id "x-ai/grok-stt-1.0". The provider exposes the model through a REST /v1/stt endpoint, so the request is an audio transcription call rather than a chat completion. Send your audio and read the transcription from the response. The supported parameters list for this model is empty, so do not assume you can pass the usual sampling or generation knobs; the features that matter here, word-level timestamps, speaker diarization, and multichannel handling, are described as part of the model's transcription behavior rather than as tunable request parameters in the data provided.

There is little in the listed facts to steer you elsewhere, but the scope is worth noting. This is a transcription-only model: audio in, transcription out. If your task is text generation, reasoning, or anything that is not speech-to-text, a different model is the right choice. If you need a documented context window or output token budget to size a request, none is given here, so verify against the live endpoint before you build assumptions into your integration.