PrismML announces Ternary Bonsai, a new family of 1.58-bit quantized language models in 8B, 4B, and 1.7B sizes. Models achieve ~9x memory reduction vs 16-bit while improving performance over 1-bit Bonsai, with the 8B variant reaching 75.5 average benchmark score. Weights released under Apache 2.0 and run natively on Apple devices via MLX.
ModelsFEATURED
Ternary Bonsai: Top Intelligence at 1.58 Bits
PrismML's 1.58-bit Ternary Bonsai models achieve 9x memory compression while outperforming their 1-bit predecessors, bringing extreme quantization and edge inference to Apple devices.
Tuesday, April 21, 2026 12:00 PM UTC2 MIN READSOURCE: Hacker NewsBY sys://pipeline
Tags
models