AssemblyAI Universal-3.6 Pro Realtime $0.45/hr vs Sume STT per minute

AssemblyAI's Universal-3.6 Pro Realtime post says the price is $0.45/hr of audio, unchanged from 3.5 Pro, with volume discounts at any tier. That is $0.0075 per minute. Sume's speech-to-text is $0.01 per audio minute, or $0.60 an hour.

AssemblyAI's price is from its launch post; Sume's from the rate card and the video inspect docs, read 2026-10-01. The AssemblyAI figure is for realtime; Sume STT is a batch job.

What is unchanged?

The post states Universal-3.6 Pro Realtime is $0.45/hr of audio, the same as 3.5 Pro. It also says volume discounts apply at any tier, without publishing the tiers. A model upgrade at the same list price means your forecast does not move because of the version change.

How does the hourly rate convert?

Dividing $0.45 by 60 gives $0.0075 a minute. The table scales both rates to common volumes.

What differs besides the rate?

The two are different shapes: AssemblyAI's is a streaming model built for voice agents, while Sume's request is a job with a 10-minute ceiling per request. Sume's duration_seconds (1 to 600) only improves the usage reservation; omitted, it reserves 1 minute.

How should I compare them for my workload?

Start from audio hours, apply each rate, then add what each shape forces on you: splitting long files into 10-minute pieces on Sume, or holding a socket open for live audio. Worked backlog math shows the pattern.