PrismML Bonsai: First Commercially Viable 1-Bit LLMs — 8B Model at 1.15GB, 14x Smaller, Apache 2.0

PrismML / Hacker News·high signal

PrismML emerged from stealth March 31 with Bonsai, claiming the first commercially viable 1-bit LLMs built on Caltech research. The 8B model fits in 1.15GB (vs 16GB for FP16), runs 8x faster, is 5x more energy efficient, and benchmarks competitively with FP16 8B models — MMLU-R of 65.7 at 1-bit precision. Ships in 8B, 4B (0.5GB), and 1.7B (0.24GB) variants under Apache 2.0, with GGUF on Hugging Face. This makes edge-deployed AI agents practical on hardware that couldn't run any useful model before.