Built for voice agents that can't wait 600ms

Qwen Alibaba Cloud: Qwen3 ASR Flash Filetrans

qwen/qwen3-asr-flash-filetrans
Speech to textactive

Qwen long-file speech recognition model served through an asynchronous transcription task API: the caller submits a file URL, polls a task id and fetches a result JSON. Accepts files up to 2 GB and 12 hours, and is the only Qwen ASR surface that returns timestamps — sentence-level always, word-level when enable_words is set. Adds multi-channel transcription (channel_id) and corpus context biasing to the Qwen3-ASR feature set. Long-file sibling of qwen3-asr-flash (synchronous batch) and qwen3-asr-flash-realtime (streaming).

ProviderSlugPriceLatencyUptime