PrismML shrinks a reasoning model to 5.9 GB, aiming for phones and PCs

PrismML hopes its tiny LLM will change how we all use AI

PrismML released Bonsai 2 27B, a compressed version of Alibaba's Qwen3.8 27B that fits into 5.9 GB — roughly a 9–10x memory reduction, small enough for PCs and possibly high-end phones. The team is led by Caltech compression expert Babak Hassibi, with Databricks co-founder Ion Stoica as an adviser. The startup raised a $22.25M seed round. Rumors of Apple talks are unconfirmed. I'd hold off on the phone hype until latency and power numbers surface.