Deepgram Flux Multilingual $0.0078/min: 7,500 minutes vs Sume STT $75

Deepgram lists Flux Multilingual streaming at $0.0078 a minute. For 7,500 minutes that is $58.50, against $75.00 on Sume STT. The full Deepgram ladder is below.

3 min readSume
All posts

Deepgram Flux Multilingual costs $0.0078 per minute on Deepgram's pay-as-you-go page, so 7,500 minutes comes to $58.50. Sume STT costs $0.01 per audio minute, so the same 7,500 minutes is $75.00. If you do not need streaming, Deepgram's pre-recorded Nova-3 Multilingual is lower still at $0.0052 per minute, or $39.00 for 7,500 minutes.

Figures were read on Deepgram's pricing page and Sume's public catalog on 2026-10-09.

The Deepgram speech-to-text ladder

Deepgram splits its price by model and by whether audio is streamed or pre-recorded. Flux is the streaming family: Flux English is $0.0065 per minute and Flux Multilingual is $0.0078 per minute. Nova-3 streaming is $0.0048 per minute for monolingual audio and $0.0058 for multilingual. Pre-recorded Nova-3 drops to $0.0043 monolingual and $0.0052 multilingual.

Deepgram also states that it bills true per-second with no rounding, so a 7,500-minute month is what you actually sent, not a count of started minutes.

7,500 audio minutes, per-minute prices read 2026-10-09
OptionPer minute7,500 minutes
Deepgram Flux Multilingual (streaming)$0.0078$58.50
Deepgram Nova-3 Multilingual (streaming)$0.0058$43.50
Deepgram Nova-3 Multilingual (pre-recorded)$0.0052$39.00
Deepgram Nova-3 Monolingual (pre-recorded)$0.0043$32.25
Sume STT 1.0$0.01$75.00

What you get for Sume's higher rate

Sume STT is an asynchronous job, not a streaming socket. You send a public HTTPS audio_url, an optional language_code, and an optional duration_seconds up to 600, and you get back text, language fields and word-level timings. Sentence segmentation is available if you add a segmentation block. The job is billed through the same ledger as Sume's other media tools.

If you already hand finished audio to Sume for captions, timeline renders or avatar videos, one vendor and one balance may be worth the premium. If you transcribe raw volume in bulk and nothing else, Deepgram's pre-recorded rate is the cheaper path on paper.

How to pick

Use the table as a quick filter, then test your own audio. Per-minute price is only half of the cost: accuracy on your languages decides how much human review you pay for afterward, and neither price list tells you that.

  • Live captions or voice agents: price the streaming rows, since a batch job cannot serve them.
  • Overnight archives: price the pre-recorded rows.
  • Audio that already lives in Sume media: the Sume row avoids moving files between vendors.

Reading the 7,500-minute figure

7,500 minutes is 125 hours, roughly the audio of a few hundred short calls or a season of a daily podcast. At that volume the gap between the cheapest and the dearest row in the table is $42.75 ($75.00 - $32.25). That is real money for a hobby project, and small next to the cost of a person fixing transcripts for a day.

So the useful question is how much cleanup each option needs on your audio, and whether you need streaming at all. Streaming rows cost more per minute than pre-recorded rows on Deepgram's own page, so a team that transcribes after the fact is paying a premium if it picks a streaming model.

Sume's catalog price is flat across languages, which keeps the estimate simple: minutes x $0.01. Deepgram's multilingual rows cost more than its monolingual rows, so a mixed-language archive should be costed language by language.

Sources

Related posts

More in Comparisons

All Comparisons posts

Written by Sume