← All models

Whisper Large v3

Audio Available

openai/whisper-large-v3 · by OpenAI · encoder-decoder-transformer

Pricing — 1 offering(s)

Groq

Audio minutes

$0.0019 / minute Current 2024-10-01 → present

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

Capability profile

transcription accuracy strong

language support strong

speaker diarisation weak

real time streaming moderate

custom vocabulary weak

Benchmarks

Benchmark	Score	Config	Source
WER improvement vs large-v2	—	10-20% relative error-rate reduction across most languages, per OpenAI's own release notes — no single absolute WER figure published for large-v3 the way whisper-1's large-v2 LibriSpeech numbers are.	source ↗
Throughput (Groq hosting)	217 x real-time	Groq's current stated throughput figure on their pricing page. An earlier Artificial Analysis benchmark (Groq's own blog, https://groq.com/blog/groq-runs-whisper-large-v3-at-a-164x-speed-factor-according-to-new-artificial-analysis-benchmark) measured 164x — Groq's hosting has been further optimised since.	source ↗

Operator guidance

Choose Whisper Large v3 on Groq when cost matters and you want a meaningfully more accurate Whisper checkpoint than OpenAI's own whisper-1 (large-v2). At $0.00185/min it undercuts every other STT option in this registry, including whisper-1 itself. Choose Nova-3 or Universal-2 instead if you need native streaming or diarization: this entry, like whisper-1, is batch/file-upload only.

Use cases

Cost-sensitive batch transcription at scale — cheapest STT option in this registry
Higher-accuracy alternative to whisper-1 for the same OpenAI Whisper lineage
Multilingual and accented-speech transcription where large-v2-based whisper-1 underperforms

Limitations

No streaming endpoint — batch/file-upload only via Groq's API
No native speaker diarization
No custom vocabulary support
10-second minimum billing per request