Qwen-Audio-3.1-TTS-Flash

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit

Not collected

Maximum input tokens

Not collected

Maximum output

Not collected

Input modalities

TextSource for Input modalities (opens in a new tab)

Output modalities

AudioSource for Output modalities (opens in a new tab)

Reasoning

Not collected

Tool calling

Not collected

Structured output

Not collected

Streaming

SupportedSource for Streaming (opens in a new tab)

Incremental synthesis/transcription or duplex audio through the documented realtime protocol.

Batch processing

Not collected

Fine-tuning

Not collected

Knowledge cutoff

Not collected

Release date

Not collected

Parameters

Not collected

Architecture

Not collected

Languages

Not collected

Context caching

Not collected

Maximum prompt characters

20,000Source for Maximum prompt characters (opens in a new tab)

Per call/segment in the SpeechSynthesizer SDK. Bidirectional synthesis permits 200,000 cumulative characters; not a token context window.

Streaming text limits

20,000 characters per segment; 200,000 cumulativeSource for Streaming text limits (opens in a new tab)

Audio formats

PCM, WAV, MP3, OpusSource for Audio formats (opens in a new tab)

Audio sample rates

8000, 16000, 22050, 24000, 44100, 48000Source for Audio sample rates (opens in a new tab)

Output sample rates in Hz; default 22,050 Hz. These are audio sampling rates, not billable token rates.

Voice controls

Volume 0–100; rate and pitch 0.5–2.0; MP3/Opus bitrate 6–510 kbpsSource for Voice controls (opens in a new tab)

Request mode

WebSocket synthesis; complete-input or bidirectional text streamingSource for Request mode (opens in a new tab)

Stock voice catalog

Use the separate Qwen-Audio-3.1-TTS system voice listSource for Stock voice catalog (opens in a new tab)

Do not reuse 3.0 voice IDs without checking compatibility.

Usage metering

Separate input and output token chargesSource for Usage metering (opens in a new tab)

Speech synthesis tokens are not interchangeable with the language-model question estimator.

Data retention

Standard API inference prompts and responses are retained up to 30 days, subject to legal/security exceptions; stateful files remain until deleted. The 3.1 audio family is listed as eligible for opt-in workspace ZDR, subject to account approval and covered paths. Eligibility does not mean ZDR is enabled by default.Source for Data retention (opens in a new tab)

Voice design

SupportedSource for Voice design (opens in a new tab)

The dedicated voice-design API lists this target model in Beijing. Earlier generic overview copy disagrees; this does not imply Singapore support.

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

Why it is included

Inference from the release-specific hosted Model Studio identity, official technical/deployment documentation and reviewed Qwen/FunAudioLLM public-weight inventories. No matching weights are published in those inventories. Public Qwen-Audio, Qwen3-TTS, Qwen3-ASR, CosyVoice and Fun-ASR checkpoints have distinct names; no equivalence with this hosted selector is assumed. This is not a claim that the whole family or all Alibaba models are closed-weight.

Sources reviewed