PrismML Compresses Qwen3.6 27B Into A Phone-Ready Model
Bonsai 27B runs natively on Apple devices and Nvidia GPUs through MLX and CUDA, in two quantized sizes.
- PrismML calls Bonsai 27B the first model of its class able to run on a phone.
- The ternary version needs 5.9 gigabytes and keeps 95 percent of Qwen3.6 27B's benchmark score.
- PrismML, founded by Caltech researchers, is backed by Khosla Ventures, Cerberus, Google and Samsung.
Why it matters: A company's claim to be first in its class is an advertisement, not a verified fact yet.
PrismML official announcement ↗ · Jul 14, 20267/14/26 · ✓ Checked✓ Check