Baseten optimizes Qwen3-ASR: 1 hour of audio transcribed in 5 seconds

Baseten engineers optimized Alibaba's Qwen3-ASR 1.7B into the fastest and cheapest batch transcription solution, processing an hour of audio in 5 seconds at about $0.0013 per hour, with 7x higher throughput than stock vLLM.

2026-10-29 ~ 2026-10-29 · 2 related posts

• Baseten makes Qwen3-ASR transcribe hour-long audio in 5 seconds at ~$0.0013/hr — baseten · 2026-10-29

1 near-duplicate retellings: baseten