Baseten makes Qwen3-ASR transcribe hour-long audio in 5 seconds at ~$0.0013/hr
baseten · x · 2026-10-29
Baseten optimized Qwen3-ASR 1.7B into the fastest, cheapest batch transcription setup: hour-long audio transcribed in under 5 seconds at about $0.0013 per audio hour, 7x stock vLLM throughput on a single GPU and a 3,128x real-time factor. They claim it's the most accurate open-weight model of its size on the Open ASR Leaderboard Pareto frontier. Key gains came from native audio decoding, KV cache quantization, and cutting unnecessary CPU-GPU syncs.
Related event: Baseten optimizes Qwen3-ASR: 1 hour of audio transcribed in 5 seconds(2 posts)→
More from Infra
- Reve publishes technical report on training infra that scaled to 10K GPUs with 10x fewer chips — cantrell · 2026-10-29
- DeepSeek Flash fingerprint change hints at migration to a new GPU cluster — teortaxesTex · 2026-10-29
- Coding agents burn 95-99% cached tokens per benchmark task, token usage chart shows — AccBalanced · 2026-10-29
- OpenRouter's 10T Daily DeepSeek Tokens Are 95% Cache Hits; One GB300 Rack Could Serve It All — AccBalanced · 2026-10-29
- Kling 4.0 video model lands on Cloudflare AI Gateway at $0.157/sec for 720p — ritakozlov · 2026-10-29
- Google dodges TPU margin question, admits AI Mode queries are monetizing poorly — firstadopter · 2026-10-29