Avencera

Pricing

Pay for audio, not seats.

Transcripts cost $0.06 an hour of audio, or $0.08 with a speaker on every line. Buy prepaid credit, with no subscription and no minimum.

Avencera prices

ModelTranscriptWith speaker labels
Parakeet v2 English$0.06 / hour$0.08 / hour
Parakeet v3 25 European languages$0.07 / hour$0.09 / hour

Compare the monthly cost

Transcript type

At 100 hours a month, Avencera costs $8.00. The next cheapest, Soniox Async, costs $9.75.

ProviderPer hourPer month
Avencera Parakeet v2 $0.08$8.00
Avencera Parakeet v3 $0.09$9.00
Soniox Async Billed by tokens, about this much per hour of speech.$0.098$9.75
Gladia Solaria-3 $0.168$16.79
AssemblyAI Universal-2 $0.17$17.00
Microsoft Azure Speech batch transcription $0.18$18.00
Rev Reverb $0.20$20.00
pyannoteAI STT Orchestration Developer plan $0.213$21.27
ElevenLabs Scribe v2 $0.22$22.00
AssemblyAI Universal-3.5 Pro $0.23$23.00
Speechmatics Melia 1 $0.24$24.00
Deepgram Nova-3 Pay as you go plan $0.258$25.80
OpenAI gpt-4o-transcribe-diarize Billed by tokens, about this much per hour. Retires February 2027.$0.36$36.00
Amazon Transcribe $0.36$36.00
Google Cloud Chirp 3 $0.96$96.00

List prices for pre-recorded audio from each provider’s pricing page, checked October 23, 2026, before free credits, discounts, or contracts. A plan with a monthly fee costs at least that fee. Euro prices are converted at $1.12. Avencera prices apply when the API opens.

Accuracy

More accurate than Whisper

Parakeet v2 averages a 6.02% word error rate on the Hugging Face Open ASR Leaderboard, against 7.64% for Whisper Large v3 Turbo and 7.16% for Whisper Large v3.

Ahead of Deepgram Nova-3

Artificial Analysis measured a 4.5% word error rate for Parakeet v3, hosted by Together AI, against 5.2% for Deepgram Nova-3 and 5.9% for Rev.

Close to OpenAI, for less

On the same leaderboard, Parakeet v2 is within 0.1 points of OpenAI gpt-transcribe, which scores 5.96% and costs over four times as much an hour.

Lower is better. Leaderboard results as of October 21, 2026, and Artificial Analysis results as of October 23, 2026.