# PrismML

[https://timesof.ai/platform/prismml](https://timesof.ai/platform/prismml)

Caltech-spinout lab building extreme on-device compression and 1-bit edge models so frontier-class intelligence can run locally on phones and laptops.

PrismML compresses large language models for local inference—backed by Khosla Ventures and exclusive Caltech IP—targeting phones, laptops, and edge devices rather than datacenter-only scale.

## Coverage

- [PrismML’s 27B-on-iPhone Claim: Edge Density vs Cloud Default](https://timesof.ai/2026/07/prismml-qwen-27b-iphone-on-device) — PrismML claims Qwen ~27B compressed under 4 GB for iPhone 15+—vendor speed/energy figures; Apple talks early; independent battery and quality tests open.