Fermion Open-Sources Phonon-2.5, a 156MB Speech Recognition Model Outperforming Whisper large-v3

Fermion AI has open-sourced Phonon-2.5, a multilingual speech recognition model that is only 156MB—about 1/20 the size of Whisper large-v3—yet achieves lower average word error rate across 25 FLEURS languages, setting a new accuracy-per-size frontier.

2026-10-24 ~ 2026-10-24 · 2 related posts

• 156MB Phonon-2.5 Beats Whisper large-v3 on Multilingual ASR at 1/20th the Size — julianweisser · 2026-10-24
• Open-source Phonon-2.5: 156MB speech model beats 20x-larger Whisper large-v3 — julianweisser · 2026-10-24