ChainThink report, April 1, 2026: According to 1M AI News monitoring, PrismML, an AI lab co-founded by mathematician Babak Hassibi from Caltech, has emerged from stealth mode and open-sourced the 1-bit Bonsai series of large language models. The flagship model, 1-bit Bonsai 8B, features 8.2 billion parameters with a memory footprint of just 1.15 GB—approximately 14x smaller than comparable 16-bit models—and concurrently released two smaller variants: 4B (0.5 GB) and 1.7B (0.24 GB).

The Bonsai 8B is a true end-to-end 1-bit model, where all layers—including embedding, attention, MLP, and output head—represent weights exclusively using +1 or -1, with no high-precision patches. PrismML asserts that its inference and language understanding performance on standard benchmarks matches that of full-precision 16-bit models. The core compression mathematics were developed over several years at Caltech and are owned by Caltech, with PrismML holding exclusive licensing rights; the model was trained on Google's v4 TPUs.

In benchmarked speed tests, the model achieves 136 tok/s on M4 Pro Mac, 440 tok/s on RTX 4090, and approximately 44 tok/s on iPhone 17 Pro Max. A standard 16-bit 8B model cannot fit into any iPhone, and energy consumption is reduced by roughly 4–5x compared to 16-bit counterparts. PrismML notes that current hardware is not optimized for 1-bit inference, and the observed speed and energy advantages primarily stem from drastically reduced memory footprint. If future hardware is specifically designed for 1-bit inference, efficiency could improve by another order of magnitude.

PrismML has completed a $16.25 million SAFE and seed round, backed by Khosla Ventures, Cerberus Capital, and Caltech. Vinod Khosla, founder of Khosla Ventures, commented on the achievement: “This is not a minor iteration—it’s a major technological breakthrough, a mathematical leap, not just another small model.”