Alibaba cut the price of AI voice tools by up to 95% in a single release

Alibaba cut the price of AI voice tools by up to 95% in a single release

Alibaba cut the price of AI voice tools by up to 95% in a single release

Story highlights

Alibaba's Qwen team released Qwen-Audio 3.1, a set of five speech models, and cut prices on its voice APIs by up to 95 per cent — speech recognition down as much as 95 per cent, text-to-speech about 70 per cent, and real-time voice around 85 per cent. The AI price war that has been reshaping text models has now arrived, sharply, in audio.

The collapse in AI prices has a new front. After months of falling costs for text models, Alibaba has taken the same axe to voice.

Its Qwen team released Qwen-Audio 3.1 — five speech models covering speech recognition, text-to-speech and real-time voice — and cut the prices of the underlying services dramatically in one move.

The Cuts

The numbers are steep. Speech-recognition pricing drops by up to 95 per cent. Text-to-speech falls by around 70 per cent. The real-time voice model, used for live conversational AI, comes down roughly 85 per cent.

A 95 per cent cut is not a discount; it is a repricing of the category. Work that was marginal at last month's rates — transcribing large audio archives, adding voice to a product at scale, running live voice agents — becomes cheap enough to treat as routine.

What Is New Besides Price

Trending Stories

This is not only a price cut with the same product behind it.

The new models add capabilities that matter for real use: speaker identification with timestamps, emotion detection, and recognition of ambient sound in the speech-understanding model, and the ability to generate speech, sound effects and background audio in a single pass in the generation model. So the offer is more capability at a fraction of the cost, which is the combination that moves a market.

Why Alibaba Is Doing This

The strategy is the same one now visible across the Chinese AI industry, and it is deliberate.

Chinese labs have competed hard on price to win developers and to blunt the advantage of Western incumbents, whose models often cost more. Driving the price of a whole category toward the floor pressures competitors who cannot subsidise as deeply, and it makes Alibaba's platform the default for anyone building voice features. This lands the same week Alibaba unveiled a new AI chip and a roadmap for far larger models — parts of one push to compete across the AI stack.

The Caution For Buyers

As with the text price war, cheap is good for buyers today and worth a moment of scepticism about tomorrow.

Prices driven this low are a mix of genuine efficiency and strategic subsidy, and from outside the two are hard to separate. Rates set to win market share can rise once the field thins, and building a product on a 95-per-cent-off price assumes that price persists. The capability gains look durable; the specific numbers are a business decision that can change.

There is also the practical point that these are Alibaba's own figures for its own services, and real-world cost depends on usage patterns the headline percentages do not capture.

What To Watch

Whether Western voice-AI providers respond with cuts of their own, which would confirm a category-wide price collapse rather than one company's move. Whether the new capabilities hold up against independent testing. And whether these introductory rates survive once developers have built on them — the point at which a price war's real terms become clear.

About the Author

Tarun Mishra is a Sub-Editor at WION. He has worked with leading outlets doing investigative journalism and covering business, global affairs, technology, space exploration etc. Hi...Read More

Trending Topics