ModelsNov 18, 2026 · 22:511 source

Reson8's New 0.5B Parameter Speech-to-Text Model Outperforms Systems 20x Its Size

From marketbrief.now

Image: marketbrief.now
AI;DR

Amsterdam startup Reson8 released Rumble-2-Flash, a 0.5B parameter speech-to-text model designed to run locally on smartphones and laptops. The 1GB model can transcribe voice notes and calls without an internet connection and outperforms transcription systems 20 times its size.

The model is free under an Apache 2.0 license and supports 45 languages. It was specifically trained on audio that typically disrupts transcription tools, such as heavy accents, street noise, bad phone lines and overlapping voices.

Read the sourceSign in to suggest

Sources

One ranked edition, every morning

The stories that matter today. Free, cancel anytime.

Telegram @aidr_today