Albert Gu's H-Net boosts byte-level modeling efficiency 3-5x

Albert Gu of Cartesia published new work on hierarchical networks (H-Net), arguing byte-level modeling without tokenizers is the next scaling step, with 3-5x efficiency gains and 10x less data matching 11T-token models.

2026-10-29 ~ 2026-10-29 · 3 related posts

• H-Net v2: Byte-Level Modeling Gets 3-5x Efficiency Gains, Matches 11T-Token Model with 10x Less Data — SonglinYang4 · 2026-10-29
• Cartesia research blog argues byte-level hierarchical networks are the next scaling step past tokenizers — SonglinYang4 · 2026-10-29
• Albert Gu's new H-Net hits 3-5x efficiency gains learning directly from raw bytes — stochasticchasm · 2026-10-29