Audio inputs and outputs with the Chat Completions API.
GPT Audio (gpt-audio) is OpenAI's first generally available audio-capable model for the Chat Completions API. Launched in January 2026, it brings native audio input and output to the familiar text-based chat completions endpoint, allowing developers to build voice-powered applications without switching to a separate Realtime API or stitching together transcription + TTS pipelines.
Unlike the Realtime models (gpt-realtime-*, which use WebRTC for low-latency speech-to-speech), gpt-audio uses the standard HTTP-based Chat Completions REST API. Developers send base64-encoded audio in the message content and receive audio back — making it straightforward to integrate into existing chat-based backends. The model features an upgraded decoder for more natural-sounding voices and improved voice consistency compared to the earlier gpt-4o-audio-preview.
| Type | Rate |
|---|---|
| Input | $2.70 / 1M tokens |
| Output | $10.80 / 1M tokens |
| Audio input | $34.56 / 1M tokens |
| Audio output | $69.12 / 1M tokens |
| Release date | January 2026 (GA)OpenRouter ↗ | verified |
| Context window | 128,000 tokensOpenRouter ↗ | to verify |
| Max output tokens | 16,384 tokensOpenRouter ↗ | to verify |
| Supported modalities | Input: text, audio; Output: text, audiodevelopers.openai.com ↗ | verified |
| License | Proprietary (OpenAI API Terms) | verified |
| Benchmark | Distribution | Score | Source |
|---|---|---|---|
Audio quality & voice naturalness OpenAI reports the gpt-audio model uses an upgraded decoder producing more natural-sounding voices with better voice consistency compared to the preview. No published benchmark scores are available. | Upgraded decoder — more natural voices vs gpt-4o-audio-preview | OpenRouter ↗ |
Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.
curl https://api.therouter.ai/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $THE_ROUTER_API_KEY" -d '{
"model": "openai/gpt-audio",
"messages": [
{"role": "user", "content": "Summarize the key points from this input."}
]
}'Send base64-encoded WAV/MP3 audio in the message content. The model can reply with both text and audio.
curl https://api.therouter.ai/v1/chat/completions \
-H "Authorization: Bearer $THEROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-audio",
"modalities": ["text", "audio"],
"audio": { "voice": "alloy", "format": "wav" },
"messages": [
{
"role": "user",
"content": [
{ "type": "text", "text": "What does this audio say?" },
{ "type": "input_audio", "input_audio": { "data": "<BASE64_AUDIO>", "format": "wav" } }
]
}
]
}'OpenAI released gpt-audio as its first generally available audio model, bringing native audio input/output to the Chat Completions REST API. The model features an upgraded decoder for more natural voices, improved voice consistency, and supports both text and audio modalities in a single API call. Priced at $2.50/M input (text) and $40/M (audio input).
gpt-audio is the GA (generally available) successor to gpt-4o-audio-preview. It uses an upgraded decoder for more natural voices with better consistency. It is also available on Azure and has a broader support ecosystem. Unlike the preview, gpt-audio is a production-grade model intended for commercial deployment.
Yes. gpt-audio is fully listed on TheRouter and available via the standard Chat Completions endpoint at api.therouter.ai/v1/chat/completions. Use model ID "openai/gpt-audio". Audio input is supported via the input_audio content part format with base64-encoded WAV/MP3 data. Set the modalities and audio parameters to request audio output.
For input, gpt-audio accepts base64-encoded audio in WAV or MP3 format. The format is specified in the input_audio content part. For output, you can request WAV format audio by setting audio.format in the request body. Supported voices include alloy, echo, fable, nova, and shimmer.
TheRouter passes through OpenAI's direct pricing with no added markup. Text input: $2.50/M tokens. Text output: $10/M tokens. Audio input: $40/M tokens. Audio output: $80/M tokens. Cache discounts may apply for repeated prompts.
| source | URL | retrieved | |
|---|---|---|---|
| Release date | OpenRouter ↗ | 2026-05-29 | verified |
| Context window | OpenRouter ↗ | 2026-05-29 | to verify |
| Max output tokens | OpenRouter ↗ | 2026-05-29 | to verify |
| Supported modalities | developers.openai.com ↗ | 2026-05-29 | verified |
| License | — | — | verified |
| Audio quality & voice naturalness | OpenRouter ↗ | 2026-05-29 | to verify |
| OpenAI launches gpt-audio — first GA audio model for Chat Completions | OpenRouter ↗ | 2026-05-29 | verified |
| How is gpt-audio different from gpt-4o-audio-preview? | openrouter.ai ↗ | 2026-05-29 | to verify |
| Can I use gpt-audio with the TheRouter API? | api.therouter.ai ↗ | 2026-05-29 | to verify |
| What audio formats does gpt-audio support for input and output? | developers.openai.com ↗ | 2026-05-29 | to verify |
| What is the pricing for gpt-audio on TheRouter? | api.therouter.ai ↗ | 2026-05-29 | to verify |