Alibaba Launches Qwen-Audio 3.1, Cutting Voice API Prices by Up to 95%

Alibaba's Qwen team launched Qwen-Audio 3.1 this week, a family of five voice models, along with price cuts of up to 95% across its audio APIs. The move resets the price floor for the entire enterprise voice-AI market, right as more companies come to depend on speech transcription and synthesis for call centers, assistants, and conversational agents.

Five models, not one

Qwen-Audio 3.1 isn't a single model but a full stack: it covers automatic speech recognition (ASR), text-to-speech (TTS), and a real-time voice model (Realtime), plus two new additions, TTS-Next and ASR-Next. ASR-Next adds multi-speaker identification with timestamps, emotion detection, and ambient sound recognition. TTS-Next, for its part, generates voice, sound effects, and background audio in a single pass, instead of requiring separate layers of audio production.

The price cuts, in detail

According to Alibaba's own announcements and independent technical coverage, the cuts aren't uniform across models: the speech-recognition API (ASR) drops by up to 95%, the text-to-speech model (TTS) drops by roughly 70%, and the real-time voice model (Realtime) drops by roughly 85% versus the previous generation of Qwen pricing.

Why this squeezes the entire voice-AI market

Providers like Deepgram, ElevenLabs, and OpenAI's own transcription offering have built business models around a price per minute or hour of processed audio that was already under pressure. With Qwen setting a significantly lower price floor — and with capabilities comparable to or better than several of its models — any enterprise voice contract signed before this announcement is likely priced above today's market rate.

What a company already using voice APIs should do

Before the next contract renewal with any voice provider — Deepgram, ElevenLabs, OpenAI, or similar — it's worth recalculating the cost per hour of processed audio using Qwen-Audio 3.1's new pricing as a benchmark, even if the final decision is to stay with the current provider for integration, compliance, or regional support reasons. The goal isn't necessarily to migrate, but to have a real reference number to negotiate with — something that didn't exist a month ago at this level of pricing aggressiveness.