Positioning and capabilities
Qwen3-ASR-Flash-Filetrans is the large-file transcription variant of Qwen3-ASR-Flash, an LLM-based multilingual automatic speech recognition model. It accepts audio and produces text, automatically identifies the spoken language, and is designed to transcribe speech in complex acoustic environments. Suitable uses include transcription of recorded meetings, interviews, and other long audio files. (help.aliyun.com)
Release date
Alibaba's release history lists this model under November 17, 2025 for its domestic release and November 18, 2025 for international deployment. Official release history. (help.aliyun.com)
Limits
The current Filetrans documentation specifies one publicly accessible audio-file URL per request, a maximum file size of 2 GB and duration of 12 hours. PCM audio must be 16 kHz; other sample rates are resampled by the service. These limits are documented for the unversioned Filetrans model. The model information page lists function calling, structured outputs, web search, context caching, batch inference and fine-tuning as unsupported; it does not specify a context-window size. Speech recognition guide; model information. (alibabacloud.com)
Pricing
The following are current official prices for the unversioned qwen3-asr-flash-filetrans model. The current pricing tables do not separately list the requested dated ID, so its present price is not independently verified. Audio input is billed by duration; text output is free. Prices were checked on November 17, 2026. (help.aliyun.com)
- Alibaba Cloud China site: Beijing: CNY 0.00022 per input-audio second; Singapore, international deployment: CNY 0.00026 per input-audio second. Official CNY pricing. (help.aliyun.com)
- Alibaba Cloud international site: Singapore, international deployment: USD 0.000035 per input-audio second; Beijing: USD 0.000032 per input-audio second. Official USD pricing. (alibabacloud.com)
These are separate site and deployment-region prices, not currency conversions.