MAI-Transcribe-2 prices end with the year: budget 500 hours a month

Microsoft's two MAI-Transcribe-2 prices both expire with the year, and its pages state no price for 2027. At 500 audio hours a month, the intro prices work out to $50 for MAI-Transcribe-2 ($0.10 per hour) and $270 for MAI-Transcribe-2-Streaming ($0.54 per hour), against $300 on Sume STT 1.0 at about $0.01 per minute (read 2026-10-08).

Budget January on a range, not a number. The one safe statement is arithmetic: Sume's rate is $0.60 per audio hour for recordings that fill whole minutes, so any Microsoft rate above that would be dearer.

What Microsoft states

The MAI-Transcribe-2 post calls $0.10 per hour "a limited-time offer until the end of the year". The streaming post calls $0.54 per hour an introductory price "through the end of the year". Neither says what happens on 1 January. If you sign a volume plan, ask Microsoft in writing.

Both are also previews or new launches: the Learn page for the streaming model says public preview, no SLA.

Sume's rounding on short files

Sume bills about one cent per audio minute rounded up per job, with a one-cent minimum. A 10-minute file is 10 cents, or $0.60 an hour. A 30-second clip is still 1 cent, which is $1.20 per audio hour if all your files are that short. A 3-minute call is 3 cents.

So the same 500 hours costs $300 as 10-minute files and $600 as 30-second clips. Average file length is the number to know before you compare.

Decision rule

Batch recordings that tolerate a vendor cloud account and a fixed intro price go to Microsoft while it lasts. Anything that must stay one HTTPS call, or that needs a stable rate card in your budget, is simpler on Sume. Neither is a live path from Sume: for live audio use the streaming model, and accept its preview status.

Contract questions to ask

If you plan to commit volume, ask for the post-2026 price in writing, whether committed-use discounts apply to the new models, and whether the model name or version can change under you.

Keep your integration thin enough to swap vendors: one function that takes an audio URL and returns text with word timings. With that seam in place, moving between Microsoft and Sume is an afternoon of work, and the January price becomes an input to a decision rather than a deadline.