← All models

Qwen3-ASR-Flash

Published Publisher Alibabaqwen3-asr-flash-2025-09-08

Overview

Qwen3-ASR-Flash is a multilingual automatic speech recognition model based on a large language model. This model ID identifies the September 8, 2025 snapshot. It takes audio as input and returns text. Alibaba’s release announcement describes automatic language identification and recognition of 11 languages, including in complex audio environments. (help.aliyun.com)

Capabilities and uses

The model is intended for non-real-time speech-to-text tasks, such as transcribing recorded speech. It supports multilingual recognition and automatic language identification. The documented API is HTTP-based and OpenAI-compatible. (help.aliyun.com)

Limitations

  • Each audio input is limited to 5 minutes and 10 MB. Speaker diarization is not supported. (help.aliyun.com)
  • The model is for non-real-time recognition rather than live audio streaming. The official model information also lists function calling, structured output, web search, context caching, batch inference and fine-tuning as unsupported. (help.aliyun.com)
  • Alibaba Cloud announced retirement of this snapshot for October 10, 2026, subject to the actual change time, and listed fun-asr as a replacement. Confirm availability before integration. (alibabacloud.com)

Release date

September 8, 2025, according to the official release announcement. (help.aliyun.com)

Pricing

The following are historical official list prices, not a guarantee of current availability or billing. Audio input was billed by duration; text output was not billed. The current pricing pages no longer list this exact snapshot. Prices were checked on November 18, 2026. (help.aliyun.com)

These regional prices are distinct offerings, not currency conversions. For current terms, consult the China-site pricing page and international pricing page.