PrismML has emerged from stealth to launch what it describes as the first commercially viable 1-bit large language model. The flagship model, 1-bit Bonsai 8B, is designed to run advanced AI locally on devices, smartphones, laptops, and embedded systems.

The model’s efficiency will allow developers to build sophisticated AI applications that execute directly on devices, reducing reliance on the cloud and unlocking a new generation of edge-first applications in robotics, wearables, and personal computing that were previously impractical.

The model uses a native 1-bit structure rather than traditional 16-bit or 32-bit architectures. According to the company, the 8-billion parameter model is 14 times smaller and eight times faster than full-precision models of the same size, while being four to five times more energy efficient. It requires 1GB of memory, compared to 16GB for standard 8B models.

PrismML’s technology is based on research from Caltech and is intended to reduce the reliance on massive datacenter infrastructure for AI inference. By running models locally, the company aims to address concerns regarding latency, power consumption, and data privacy. The 1-bit Bonsai models are being released under the Apache 2.0 license.

“We are creating a new paradigm for AI: one that adapts to diverse hardware environments and delivers maximum intelligence per unit of compute and energy,” said Babak Hassibi, CEO and Founder of PrismML.

The company is also releasing smaller versions of the model, including 4B and 1.7B variants with memory footprints of 0.5GB and 0.24GB, respectively. PrismML is backed by Khosla Ventures and Cerberus Ventures, with the technology expected to influence future AI hardware and computer architecture design.

More information can be found at: PrismML — PrismML Launches World’s First 1-Bit AI Model to Redefine Intelligence at the Edge.