Qwen: Qwen-Audio-3.0-TTS Flash

Qwen-Audio-3.0-TTS Flash is a text-to-speech model from Qwen, Alibaba's model family. It sits in the speech category and is described by the provider as a fast, cost-efficient text-to-speech system. The model was released by the provider on 2026-07-23. If you are already working within the Qwen ecosystem, or you need a hosted TTS option that fits an OpenAI-style client, this is the model to evaluate.

The model takes text as its only input and produces speech as its only output. It does not accept images, audio or other media on the way in, and it does not return text on the way out. Two voices are available: loongjohn and longanhuan_v3.6. There is no published context length for this model, and the maximum output value listed is 0, so treat the text you send as the thing being spoken rather than as a long document to be reasoned over. No additional request parameters are listed as supported.

The practical uses follow directly from that shape. Narration for videos, spoken versions of articles or documentation, voice prompts for phone trees and IVR systems, audio for accessibility features, and pre-generated speech for apps that need consistent playback are all reasonable fits. Because the model is positioned as a flash variant, it suits workloads where you generate a lot of short or medium utterances and want the synthesis step to stay out of the way. The two named voices let you pick a consistent speaker per product or per locale without building your own voice pipeline.

You call it through api14's OpenAI-compatible API using the model id "qwen/qwen-audio-3.0-tts-flash". Point your existing OpenAI client at the api14 base URL, send a request to the speech endpoint with the text you want spoken, and select one of the two voices. Because the API is OpenAI-compatible, you can keep the same client library and request structure you already use for other models on the gateway, which keeps integration work small if you are adding speech to an existing application.

Keep the scope in mind when choosing. This is a speech model, so it will not help with chat, reasoning, classification or any task that expects text back. If you need a model that reads and writes text, or one that handles audio as input rather than output, a different model on api14 will fit better. Within its lane, though, it is a straightforward way to turn text into audio without leaving the OpenAI-compatible surface you already use.