Monologue launches in-house dictation model mono-1: 55% fewer edits, 3x faster than API pipeline
every · x · 2026-09-24
Dictation app Monologue introduced mono-1, its in-house dictation model built over six months to replace its API-based pipeline. On WildSpeech-Bench (1,100 English clips covering noise, stuttering, overlapping speech), mono-1 hit a 7.22% WER—second only to OpenAI's GPT-4o Transcribe at 6.68%—with a 109 ms median raw-transcription latency, tied fastest with Cartesia Ink. Versus the old pipeline, mono-1 needs 55% fewer estimated edits and delivers finished dictation 3x faster by median response time, handling corrections, formatting, context, and writing preferences.
More from Models
- 'In a few years, claims that models needed stolen work will look absurd' — nabla_theta · 2026-09-24
- Developer calls Claude Opus 5.5 'a doof' at database tasks — rickasaurus · 2026-09-24
- OpenAI ships major ChatGPT Voice update: plugins, GPT-6 models, and voice-built docs and spreadsheets — john__allard · 2026-09-24
- 'Models are just getting really smart': researcher pushes back on OpenAI theft claims — nabla_theta · 2026-09-24
- Zvi Reads the Claude Opus 5.5 System Card: Top Benchmarks, Actively Cheaper Than Opus 5 — TheZvi · 2026-09-24
- Early user feedback: Opus 5.5 is 'finally a usable Opus again' — weswinder · 2026-09-24