Build My Rig
OpenAI
Transcription810M
Transcription

Whisper Large V3 Turbo

Whisper · MIT

Distilled version of Whisper Large V3. 8x faster than the full model with minimal quality loss. The go-to for real-time transcription on local hardware. Runs comfortably on any GPU with 2GB+ VRAM.

Parameters
810M
Architecture
Dense
Context
1,500 tokens
Released
2024-10-01
Engines
whisper.cpp, faster-whisper, transformers

Parameters

810M

VRAM

1.6 GB

Context

1K

Formats

1

GPUs

43

Whisper Large V3 Turbo (810M) requires 1.6 GB VRAM at recommended quality (FP16). On NVIDIA Grace Blackwell Ultra GB300, expect approximately 600 tok/s at FP16. For the best experience, Starter AI Desktop ($582) is recommended.

Source: OwnRig methodology

VRAM (Recommended)

1.6 GB

Quantization

FP16

File Size

1.5 GB

Max Context

1K tokens

Primary Use

Transcription

Compatible GPUs

43 devices
NVIDIA Grace Blackwell Ultra GB300FP16600 tok/sExcellent
Apple M4 Max (128GB Unified)FP16–Excellent
Apple M4 Max (36GB Unified)FP16–Excellent
Apple M4 Max (64GB Unified)FP16–Excellent
Apple M4 Pro (24GB Unified)FP16–Excellent
Apple M4 Pro (48GB)FP16–Excellent
Apple M4 Ultra (192GB)FP16–Excellent
NVIDIA GeForce RTX 3060 12GBFP16–Excellent
NVIDIA GeForce RTX 3080 10GBFP16–Excellent
NVIDIA GeForce RTX 3090FP16–Excellent
NVIDIA GeForce RTX 4060 8GBFP16–Excellent
NVIDIA RTX 4060 Laptop (40-60W)FP16–Excellent
NVIDIA GeForce RTX 4060 Ti 16GBFP16–Excellent
NVIDIA RTX 4070 Laptop (80-115W)FP16–Excellent
NVIDIA GeForce RTX 4070 SuperFP16–Excellent
NVIDIA GeForce RTX 4070 Ti 12GBFP16–Excellent
NVIDIA GeForce RTX 4070 Ti SuperFP16–Excellent
NVIDIA RTX 4080 Laptop (120-150W)FP16–Excellent
NVIDIA GeForce RTX 4080 SuperFP16–Excellent
NVIDIA GeForce RTX 4090FP16–Excellent
NVIDIA RTX 4090 Laptop (150-175W)FP16–Excellent
NVIDIA GeForce RTX 5080FP16–Excellent
NVIDIA GeForce RTX 5090FP16–Excellent
AMD Radeon Pro W7900FP16–Excellent
NVIDIA RTX PRO 6000 BlackwellFP16–Excellent
NVIDIA RTX PRO 6000 Blackwell Max-QFP16–Excellent
NVIDIA GeForce RTX 5060 8GBFP16–Excellent
NVIDIA GeForce RTX 5060 Ti 16GBFP16–Excellent
Apple M3 Pro (18GB Unified)FP16–Good
Apple M4 (16GB Unified)FP16–Good
AMD Radeon RX 7600FP16–Good
AMD Radeon RX 7900 XTXFP16–Good
AMD Radeon RX 9070FP16–Good
Apple M1 (8GB Unified)FP16–Good
Apple M1 (16GB Unified)FP16–Good
Apple M1 Pro (16GB Unified)FP16–Good
Apple M2 (8GB Unified)FP16–Good
Apple M2 (16GB Unified)FP16–Good
Apple M2 Pro (16GB Unified)FP16–Good
Apple M3 (8GB Unified)FP16–Good
Apple M3 (16GB Unified)FP16–Good
AMD Radeon RX 9060 XT 16GBFP16–Good
AMD Radeon RX 9060 XT 8GBFP16–Good

Showing 43 of 43 entries

FAQ

Frequently Asked Questions

How much VRAM does Whisper Large V3 Turbo need?
Whisper Large V3 Turbo requires 1.6 GB VRAM at recommended quality (FP16).
What is the best GPU for Whisper Large V3 Turbo?
The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Whisper Large V3 Turbo, achieving 600 tok/s at FP16 with an excellent rating.
Can I run Whisper Large V3 Turbo on an RTX 4060 Ti?
Yes. On the NVIDIA GeForce RTX 4060 Ti 16GB, Whisper Large V3 Turbo runs at – (FP16, excellent).

Related Guides

Tutorial

Running Whisper locally: GPU requirements and setup

Whisper Large V3 and V3 Turbo GPU requirements, VRAM usage, and hardware recommendations. Any GPU with 4 GB handles it; here is what you actually need for production use.

Buying Guide

Mac Mini M4 for AI: which models run on 16 GB

Which AI models run on the Mac Mini M4 with 16 GB, 24 GB, or 48 GB of unified memory. Honest compatibility table, real quantization requirements, and the upgrade case for M4 Pro.

All models

Data confidence: estimated. Source

VRAM requirements are calculated from model parameters and may vary by inference engine, context length, and batch size. Performance estimates are based on community benchmarks and should be verified for your specific configuration.Whisper is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.