GPT-4o mini Transcribe

Overview

This is OpenAI’s speech-to-text engine, designed to convert audio inputs into text. It handles conversational speech, background noise, and mixed-language audio streams with high accuracy. The model accepts audio input and returns text, with no support for vision, reasoning, or structured outputs. On Kyma, it runs through an OpenAI-compatible endpoint using a single API key. Requests benefit from automatic failover, routing around degraded serving paths transparently. The platform reports exact request costs in the usage object and identifies the executed model via the X-Kyma-Model header. This model does not support prompt caching. The model operates on a per-minute billing basis and is optimized for fast processing. It does not generate audio or structured JSON, and its context window is limited to 1500 tokens. It returns a plain transcript with no per-segment timestamps, so verbose_json, srt and vtt are refused on this SKU (400 segments_unsupported).

Specs

Field	Value
Model ID	`gpt-4o-mini-transcribe-2025-12-15`
Best for	High-accuracy dictation, multilingual transcription, conversational audio
Context window	1,500
Input modalities	Audio
Output modalities	Text
Speed	Fast
Cost band	Balanced
Release stage	Stable

Pricing

	Price
Per minute of audio	$0.00405

Billed per minute of audio. Every request reports its exact cost in usage.cost; the full media price list is on Pricing.

Use this when

Conversational Audio Transcription — Captures multi-speaker dialogue and natural speech patterns with high accuracy.
Noisy Background Dictation — Filters background interference to produce clean text from field recordings or calls.
Mixed Language Audio Processing — Handles code-switched audio streams like Vietnamese and English in a single pass.

Not ideal for

Do not use this model for generating audio, running reasoning tasks, or processing files that exceed its 1500-token context limit. It is also not the model to start new work on: OpenAI shuts the gpt-4o-transcribe family down on 2027-02-26.

Pick something else when

You are starting new work, or want the model OpenAI names as this one’s replacement: use gpt-transcribe.
You need a lower-cost option for clear studio recordings: use whisper-v3-turbo.

Example

from openai import OpenAI

client = OpenAI(base_url="https://kymaapi.com/v1", api_key="kyma-...")

response = client.audio.transcriptions.create(
    model="gpt-4o-mini-transcribe-2025-12-15",
    file=open("audio.wav", "rb"),
)

Agent query example

Ask the API which models fit, instead of hardcoding an id:

curl "https://kymaapi.com/v1/models?recommended_for=transcribe-audio&quality_tier=frontier-open"

FAQ

Does this model support structured JSON output? No, it only returns raw text transcriptions from audio input. How does Kyma handle request routing for this model? Every request uses automatic failover, so if a serving path degrades, your call is rerouted without manual intervention. Can I cache prompts to reduce costs? No, this model does not support prompt caching. Is this model being retired? Yes. OpenAI announced on 2026-08-26 that the gpt-4o-transcribe family shuts down on 2027-02-26, and names gpt-transcribe as the replacement. Kyma keeps serving this id until that date; the catalogue carries the date in the retires_on field so you can plan the move.

Alias	Resolves to
`transcribe-quality`	`gpt-4o-mini-transcribe-2025-12-15`

​Overview

​Specs

​Pricing

​Use this when

​Not ideal for

​Pick something else when

​Example

​Agent query example

​FAQ

​Related aliases