Qwen-Audio 3.1 Transcribes an Hour of Calls for About Two Cents
Qwen-Audio 3.1 bills speech recognition in tokens: at 25 tokens per audio second it works out to about $0.02 an hour in Singapore, against $0.15-$0.26 at AssemblyAI and Deepgram. The catches are a five-minute request cap, per-channel billing and a residency decision.
September 27, 2026 · 10 min read