# GPT-4o mini Transcribe

GPT-4o mini Transcribe applies GPT-4o mini's audio understanding to transcription. You get more accurate transcripts than Whisper-generation models, priced for high-volume audio pipelines.

- **Model ID:** `openai/gpt-4o-mini-transcribe`
- **Type:** transcription
- **Providers:** openai
- **Context window:** Not applicable
- **Maximum output tokens:** 0
- **Pricing:** $1.25/1M input tokens, $5/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/gpt-4o-mini-transcribe

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { experimental_transcribe as transcribe } from 'ai';
import { gateway } from '@ai-sdk/gateway';
import { readFile } from 'node:fs/promises';

const result = await transcribe({
  model: gateway.transcriptionModel('openai/gpt-4o-mini-transcribe'),
  audio: await readFile('audio.mp3'),
});
```

## About

GPT-4o mini Transcribe launched on March 13, 2024 as part of OpenAI's next-generation audio models for the API. The model builds on the GPT-4o mini architecture and was pretrained on specialized audio-centric datasets. OpenAI used enhanced distillation techniques to transfer knowledge from larger audio models into this smaller one.

Compared to the original Whisper models, GPT-4o mini Transcribe improves word error rate and language recognition. OpenAI attributes the gains to reinforcement learning work and midtraining on diverse, high-quality audio data. The improvements show up most in difficult conditions: accents, noisy environments, and varying speech speeds.

Through AI Gateway, GPT-4o mini Transcribe handles transcription jobs behind the same authentication, observability, and spend controls as your text models. Send audio with the AI SDK's `transcribe` function and get back the transcript, plus segments and language metadata where available.

## What to consider

Audio support on AI Gateway is in beta. You call GPT-4o mini Transcribe with the AI SDK's `transcribe` function, passing audio as a file buffer, base64 string, or URL. If accuracy on your hardest recordings is the deciding factor, compare results against `gpt-4o-transcribe` on your own audio before settling on the mini variant.

## When to use

### Best For

- High-volume transcription pipelines where per-request cost shapes the architecture
- Call, meeting, and voicemail transcription with fewer errors than Whisper-generation models
- Products handling accented speech or imperfect recordings that trip up older models
- Adding speech input to apps that already route text models through AI Gateway

### Consider Alternatives When

- Accuracy on difficult audio is the top priority and `gpt-4o-transcribe` justifies its higher rate
- You need speech translation to English or language identification, which `whisper-1` handles as a multitask model
- Your product needs live speech-to-speech conversation rather than transcription; look at the gpt-realtime family

## Best for

- **High-volume transcription pipelines:** Per-request cost shapes the architecture and the mini rate keeps unit economics workable
- **Call and meeting transcription:** Accuracy improvements over Whisper-generation models reduce cleanup work downstream
- **Difficult audio conditions:** Accents, background noise, and varying speech speeds that trip up older models
- **Unified gateway workflows:** Adding speech input to apps that already route text models through AI Gateway

## Consider alternatives

- **Maximum transcription accuracy:** `gpt-4o-transcribe` is the stronger variant when difficult audio justifies a higher rate
- **Speech translation needs:** `whisper-1` handles translation to English and language identification as a multitask model
- **Live voice conversations:** The gpt-realtime family serves speech-to-speech agents rather than transcription jobs

## Frequently asked questions

### How do I access GPT-4o mini Transcribe through AI Gateway?

Use the AI SDK's `transcribe` function from the `ai` package with this model's slug. Audio support on AI Gateway is in beta. Authenticate with an AI Gateway API key or OIDC token; no OpenAI credentials are needed in your application.

### How does GPT-4o mini Transcribe differ from gpt-4o-transcribe?

Both are speech-to-text models from the same next-generation audio release. GPT-4o Transcribe builds on the larger GPT-4o architecture for the strongest accuracy, while GPT-4o mini Transcribe builds on GPT-4o mini and targets cost-efficient, high-volume transcription.

### Is GPT-4o mini Transcribe more accurate than Whisper?

Yes. OpenAI reports improvements to word error rate and better language recognition and accuracy compared to the original Whisper models, with the largest gains on accents, noisy environments, and varying speech speeds.

### What audio inputs does the AI SDK accept?

The `transcribe` function accepts a file buffer, a base64-encoded string, or a URL. The result includes the transcript text, plus segments, detected language, and duration where available.

### How is GPT-4o mini Transcribe priced?

Pricing is listed on this page. Rates reflect the providers routing through AI Gateway and update when providers change list prices.

### Does GPT-4o mini Transcribe support zero data retention through AI Gateway?

Zero Data Retention is not currently available for this model. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

### What are typical latency characteristics?

This page shows live performance metrics measured across real AI Gateway traffic.

## Links

- [Model page](https://vercel.com/ai-gateway/models/gpt-4o-mini-transcribe)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://developers.openai.com/api/docs/models/gpt-4o-mini-transcribe)