Skip to content

Benchmark

Speech Model Quantisation Benchmark

A repeatable comparison of quantised speech recognition models on consumer hardware, measuring word error rate against latency and memory footprint.

Status
Benchmark
Last updated
April 2026

01

So far

Aggregate word error rate hides the failure that matters. Accuracy on conversational speech survives aggressive quantisation far better than accuracy on domain vocabulary, so a single headline number will lead you to ship the wrong configuration into a clinical setting.

02

Stack

  • ONNX Runtime
  • CTranslate2
  • Whisper
  • Python

Contact

If you are working on something where being wrong matters, I would like to hear about it.

I am open to consulting engagements, research collaborations, and conversations that do not have a clear outcome yet.