Whisper Large v3.
OpenAI's open-weight speech-to-text — the standard transcription model.
1× Nvidia Titan V.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
| Provider | Access | $/M in | $/M out | |
|---|---|---|---|---|
| OpenAI Whisper API | api direct | — | — | Launch ↗ |
| Groq | hosted inference | — | — | |
| Replicate | hosted inference | — | — |
Smaller models distilled from Whisper Large v3.
Lightweight student models trained to mimic Whisper Large v3's outputs.
Whisper Medium
769M Whisper variant — half the size of Large, 80% of the accuracy.
Whisper Small
244M Whisper — fits on edge GPUs and CPU.
Whisper Base
74M Whisper — browser / Raspberry Pi-deployable.
Whisper Tiny
39M Whisper — runs in-browser via WebGPU.
Variants in the Whisper family.
Whisper Medium
769M Whisper variant — half the size of Large, 80% of the accuracy.
Whisper Small
244M Whisper — fits on edge GPUs and CPU.
Whisper Base
74M Whisper — browser / Raspberry Pi-deployable.
Whisper Tiny
39M Whisper — runs in-browser via WebGPU.
Frequently asked.
How do I run Whisper Large v3?
Where can I access Whisper Large v3?
How much does it cost to run Whisper Large v3?
Is Whisper Large v3 open-source or proprietary?
Cheapest hardware per quantisation.
Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr refreshed hourly from each provider's API.
| Quantisation | Cheapest GPU config | Total VRAM | Live $/hr | tokens/sec | |
|---|---|---|---|---|---|
|
FP16
FP16 — half precision (default)
|
4 GB | — | — | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
3 GB | — | — | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
3 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Whisper Large v3 across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
No priced API access rows on file for Whisper Large v3 yet.
Rent the GPU instead of paying per token.
For an open-weights model like Whisper Large v3, you can rent a GPU and serve inference yourself. The math: cheapest GPU rental × 730 hours/month + your electricity rate × power draw.
Assumes the GPU runs 24/7 at ~85% utilisation. If your traffic is bursty, you'll pay less for the API and probably more for the GPU (idle hours still cost rental). The breakeven analysis lives on the Self-host vs API breakeven tool.
About Whisper Large v3.
Whisper Large v3 is OpenAI's open-weight speech-to-text model — the de facto standard for automatic transcription. Supports 99 languages, runs in real-time on a single GPU. The "large-v3" 2023 release added 5% accuracy gains over v2 plus better noise handling. Most popular transcription model on Hugging Face by a wide margin.