Inference APIs
Compare/Speech-to-text

Parakeet TDT 0.6B v3 vs Voxtral Small 24B

Two open-weight speech-to-text models on the same endpoint and API key, so switching between them is a one-word change. Below: what each costs for three realistic workloads, how they differ, and measured speed.

Inference APIs prices are read live from the price list.

Short answer
  • Cost: Parakeet TDT 0.6B v3 is cheaper in all three workloads, by 50%.

Side by side

Parakeet TDT 0.6B v3Voxtral Small 24B
Served byInference APIsInference APIs
Model authorNVIDIAMistral AI
Model idnvidia/parakeet-tdt-0.6b-v3mistralai/Voxtral-Small-24B-2507
Price$0.002 / audio minute$0.004 / audio minute
CapabilitiesSpeech to text, Very fast, Timestamps, 25 languagesSpeech to text, 8 languages, Auto language, Text only
WeightsOpenOpen
StatusAvailableAvailable
LimitsNo per-minute or per-day request or token caps; usage draws on a prepaid balanceNo per-minute or per-day request or token caps; usage draws on a prepaid balance
Speed0.036× real time (19.5 s clip in 0.7 s)0.09× real time (23.1 s clip in 2.08 s)

What three workloads cost

List prices applied to the same work. The cheaper side of each row is in bold.

WorkloadParakeet TDT 0.6B v3Voxtral Small 24BDifference
1 hour of audio$0.12$0.2450%
100 hours of audio$12.00$24.0050%
1,000 hours of audio$120$24050%

Parakeet TDT 0.6B v3 is cheaper in all three workloads, by 50%.

When to choose which

Choose Parakeet TDT 0.6B v3 if
  • Cost is the deciding factor; it is cheaper in every workload above
  • Long recordings in one of its 25 European languages, where throughput matters
  • You need timestamps or subtitles (use Whisper or Parakeet)
  • Languages outside its eight; use Whisper
  • Speaker diarization
Choose Voxtral Small 24B if
  • Accuracy on messy real-world audio in English, Spanish, French, Portuguese, Hindi, German, Dutch or Italian
  • You want Mistral's Voxtral without running a 24B model yourself
  • Your audio is in a language outside its 25; use Whisper
  • You need live streaming transcription or speaker diarization

Measured speed

Median of three runs on 2026-09-16: a 19.5-second mp3 clip, timed from upload to response. Expect ±30% with time of day. Method and raw numbers: speed measurements.

Trying both

Same endpoint, same key. Change one string:

cURL
for MODEL in "nvidia/parakeet-tdt-0.6b-v3" "mistralai/Voxtral-Small-24B-2507"; do
  curl -s https://api.inferenceapis.com/v1/audio/transcriptions \
    -H "Authorization: Bearer $INFERENCE_API_KEY" \
    -F "model=$MODEL" -F "file=@meeting.mp3" | jq -r .text
done

FAQ

Which is cheaper, Parakeet TDT 0.6B v3 or Voxtral Small 24B?

Parakeet TDT 0.6B v3 is cheaper in all three workloads, by 50%. The table above uses list prices: Parakeet TDT 0.6B v3 $0.002 / audio minute; Voxtral Small 24B $0.004 / audio minute.

How do I switch between them?

They are on the same endpoint and key. Change the model field from nvidia/parakeet-tdt-0.6b-v3 to mistralai/Voxtral-Small-24B-2507 and nothing else.

Parakeet TDT 0.6B v3 API Voxtral Small 24B API Playground

Spotted a price or limit that has changed? Tell us and we will re-check the provider page.