Official documentation
open-audio-opd: Audio OPD training and ARK-ASR inference (README)
AutoArk, GitHub, AutoArk repository, 2026
Read the originalRead on October 21, 2026
- What it is
- The README of AutoArk's open-source repository for online policy distillation of speech models. Its Model Release table names the ARK-ASR-0.6B checkpoint, its task and the languages it recognizes.
- How much weight it carries
- The maker's own repository documentation, dated by its announcements (2026.05.25 weights, 2026.05.28 demo). It states the language list but not how the list was tested, and it names only the 0.6B checkpoint.
Facts taken from this source
- AutoArk's open-audio-opd repository lists 19 languages for its ARK-ASR-0.6B speech recognition checkpoint, from Chinese and English to Lithuanian.
“Task Autoregressive ASR Languages Chinese, English, German, Japanese, French, Korean, Spanish, Polish, Italian, Romanian, Hungarian, Czech, Dutch, Finnish, Croatian, Slovak, Slovene, Estonian, Lithuanian” (Model Release table)