Model catalog

The cheapest transcription APIs in 2026

Transcription on the catalog starts at $0.0015 per minute of audio (Deepgram Flux). At that rate 1,000 hours of audio, 60,000 minutes, transcribes for $90 at list. All 7 transcription models are compared below.

01
Deepgram · 0K context · $0.0015 per minute · Proprietary API
02
Deepgram · 0K context · $0.0015 per minute · Proprietary API
03
Deepgram · 0K context · $0.0015 per minute · Proprietary API
04
Nvidia · 0K context · $0.0015 per minute · Apache 2.0
05
Nvidia · 0K context · $0.0015 per minute · Apache 2.0
06
Nvidia · 0K context · $0.0015 per minute · Not listed
07
OpenAI · $0.0015 per minute · Apache 2.0

Provider list prices from the Allocate catalog, checked 2026-07-08. Billed price on Allocate is list plus the 7% transaction fee.

Per-minute pricing in practice

Transcription bills on audio duration, not words produced, so cost forecasting is simple: minutes in, dollars out. Streaming and batch modes of the same model usually share a price; latency requirements, language coverage, and diarization decide between models more than price does at these levels.

For downstream LLM work, remember the second bill: an hour of speech is roughly 9,000 words, about 12,000 tokens, of input to whatever model reads the transcript.

Common questions

What does transcription cost per hour?

$0.09 per hour at list on Deepgram Flux ($0.0015 per minute). On Allocate, list plus the 7% transaction fee.

Does audio quality change the price?

No; billing is per minute of input audio regardless of quality. It changes accuracy instead, which is why noisy-channel workloads are worth testing across models before committing volume.

Related

Every model here sits behind one key on Allocate: route by name, meter per route, and swap the model in one click.