Higurashi Open-Sources Haru 4B Full-Duplex Speech Model

Higurashi AI open-sourced Haru 4B, an end-to-end full-duplex speech dialogue model trained on 40 million hours of multilingual data, released under Apache 2.0 with weights and inference code.

2026-11-13 ~ 2026-11-13 · 2 related posts

• Higurashi open-sources 4B full-duplex voice model that laughs and nods naturally — HigurashiAI · 2026-11-13
• Higurashi open-sources Haru 4B, a full-duplex speech model trained on 40M hours — alexcovo_eth · 2026-11-13