PrismML releases compressed AI model small enough to run on smartphones and PCs
PrismML’s Bonsai 2 27B cuts Alibaba’s Qwen3.8 27B to 5.9 GB—9–10x less memory—while matching 98% of its aggregate benchmark scores. Ternary weights use only +1, −1, or 0, enabling local PC and potentially smartphone deployment.
AI startup PrismML released Bonsai 2 27B on Thursday, September 18, 2026 — a compressed reasoning model that shrinks Alibaba’s open-source Qwen3.8 27B down to 5.9 GB, small enough to run on a PC and potentially a high-end smartphone. That represents a 9x to 10x reduction in memory compared to the original model.
PrismML achieves the compression using what it calls “ternary” weights. Standard model weights require 16 bits of storage each; PrismML’s approach reduces each weight to one of just three values — +1, −1, or 0 — dramatically cutting the space required. Despite that reduction, Bonsai 2 matches 98% of Qwen’s aggregate benchmark scores, up from 95% achieved by the first Bonsai model released in March 2026.
The company was founded by a group of Caltech researchers and is led by CEO Babak Hassibi, a Caltech professor specializing in compression technologies. Ion Stoica, co-founder of Databricks and director of Berkeley’s Sky Computing Lab, serves as an adviser. Investors include Khosla Ventures, Cerberus Capital, and Caltech. The startup has raised a $22.25 million seed round to date.
PrismML’s earlier models have already seen significant uptake — the original Bonsai has been downloaded over 11 million times, with its smaller models adding another 2.6 million downloads, according to the company. The startup is also rumored to be in talks with Apple, though Hassibi declined to comment on that.
Hassibi told TechCrunch that PrismML’s next releases, expected within a few months, will target models in the several-hundred-billion-parameter range, and he expects compression to be more effective at that scale. “As a general trend, for larger models, it’s easier to get to 100%,” he said.
Stoica framed the technology’s significance in terms of user access and privacy. “You are going to have intelligence at your fingertips, and it’s going to be free because it’s going to run on the device you already bought,” he said. “It’s also going to be private, because you’re not going to send it to the cloud.”