the-decoder.com web signal

Alibaba Ships Qwen-Audio 3.1 Stack, Cuts Voice APIs Up to 95%

By Published Updated
6 sources tracking this story
Alibaba Voice AI China AI ai-business

TL;DR

  • A 95% ASR price cut resets the cost baseline the entire transcription vendor category must meet, regardless of what prior margins were built on.
  • TTS-Next generates voice, sound effects, and background audio in a single pass, collapsing three separately orchestrated inference systems into one model call.
  • Realtime Plus's 262K token context lets a voice agent carry conversation history, tool results, and application state across a full session without round-tripping three separate systems.

Alibaba's Qwen team pushed a new voice stack and, in the same post, cut prices on its audio APIs by up to 95 percent. The-Decoder reports that ASR drops by up to 95 percent, TTS by about 70 percent, and Realtime by roughly 85 percent.

The lineup is five models: upgraded versions of ASR, TTS and Realtime, plus two additions, TTS-Next for audio creation and ASR-Next for audio understanding.

The capability descriptions are specific. The upgraded ASR "automatically cleans up filler words and repetitions" and improves multilingual and dialect recognition. ASR-Next layers in multi-speaker identification with timestamps, emotion detection, and ambient sound and machine noise detection. TTS delivers "multilingual synthesis with natural cross-language voice transfer," with emotion, speed and style set through text prompts. TTS-Next generates "voice, sound effects, and background audio in a single pass." The Realtime model handles simultaneous speaking and listening with "instant interruption."

The drop lands in a busy stretch of Alibaba AI coverage, and slots alongside the other new voice releases moving through our Voice AI feed this week from Google and OpenAI. The article does not include an effective date for the discounted pricing or say whether the tier is capacity-committed or spot.

What others are reporting

Coverage cluster as of 8h after publish

  1. Metaverse Post Read →

    Places Qwen-Audio 3.1 inside Alibaba's full vertical play: Zhenwu V900 silicon, 5-10T parameter Qwen 4/5 roadmap, 20GW cloud target, and revenue-sharing license terms at scale.

    TTS costs dropped roughly 70%, Realtime around 85%, and ASR by up to 95%.
  2. AlphaSignal Read →

    Developer-focused breakdown of TTS-Next's LM-plus-diffusion architecture with practical guidance on turn latency, context consumption, and total cost-of-ownership for teams evaluating a switch.

    Fewer orchestration boundaries can reduce latency and simplify interruption handling, especially when a user speaks over the agent.
  3. Data Studios Read →

    Centers on the stateful agent runtime case: 262K context sustains full session state including tool results across a conversation, repositioning voice AI from chatbot to agent infrastructure.

    A single realtime model can interpret speech and generate spoken responses without forcing every turn through three separately orchestrated inference systems.
  4. Orca Router Read →

    Direct pricing comparison showing 70% undercut versus ElevenLabs at $100/M characters, with 20 Chinese dialect regions ElevenLabs lacks, but only 16 languages versus ElevenLabs' 74.

    A 70% cut against a $100-per-million-character incumbent is not a rounding difference. It is the difference between a product you prototype with and a product you ship.
  5. TBreak Read →

    Notes ASR-Next adds multi-speaker timestamps, emotion detection, and ambient sound recognition; flags absence of UAE-specific pricing as a gap for regional developers.

    Lower per-call costs reduce the pressure to limit usage.

Originally reported by the-decoder.com

Read the original article →

Original headline: Alibaba Ships Qwen-Audio 3.1 Stack With TTS-Next and ASR-Next, Cuts Voice API Prices Up to 95%