OpenAI: GPT Transcribe

GPT Transcribe is a speech-to-text model from OpenAI, listed on api14 under the model id openai/gpt-transcribe. It sits in the transcription category, so its job is narrow and specific: turning audio into text rather than generating conversational replies or handling general purpose chat. If you are building a pipeline where spoken content needs to become written content, this is the kind of model the provider intends for that step.

The model takes audio as its input modality and returns a transcription as its output. According to the provider description, it is a high accuracy speech-to-text model suited to recorded audio, streamed file transcription, and committed Realtime turns. It supports free-form context, keyword hints, and multiple languages. No context length is published for this model, and the maximum output is listed as zero, so you should treat the practical limits as something to verify against the provider's own documentation rather than assume a token budget from this page. What the page does make clear is the shape of the interaction: audio in, text out, with optional context and keyword guidance to steer recognition.

The use cases follow directly from that. Recorded audio transcription covers meeting recordings, interviews, podcasts, voice memos and any archive of spoken material you want searchable as text. Streamed file transcription fits workflows where audio arrives as a file and you want the transcript produced as part of a larger processing job. Committed Realtime turns point to live or near live speech being transcribed as part of an interactive session. Free-form context is useful when you can tell the model what the audio is about, for example domain vocabulary or the subject of a call, and keyword hints help when you know specific terms, names or jargon are likely to appear. Multiple language support means the same endpoint can serve multilingual audio rather than forcing you into a separate model per language.

You call it through api14's OpenAI-compatible API using the model id openai/gpt-transcribe, so existing OpenAI client libraries and request patterns should carry over with the base URL pointed at api14. The supported parameters list for this model is empty, which means there are no additional request parameters documented here beyond what the endpoint accepts by default. Practically, that means you should send your audio and the model id, and not rely on tuning options such as temperature or custom sampling settings appearing on this page. If you need a parameter that is not listed, check the provider's documentation before assuming it is available.

Because the model is transcription only, it is not a fit for tasks that need text generation, reasoning, tool calling or image input. If your workload is chat, code or general text, a different model in the catalog will serve you better. There is also no published context length and a maximum output of zero on this page, so for very long audio or unusual throughput requirements you should confirm behaviour with the provider rather than infer it from these fields.