Bonsai 2 2B: a 911 MB Ollama download we ran on a CPU, no GPU
Bonsai 2 is from PrismML, free under Apache 2.0, and every weight is -1, 0 or +1. In Ollama, bonsai-2:2b is a 911 MB download, and it reads images. Our test, on a CPU-only box with 8 cores and no GPU: it wrote a heater text for a landlord and got 17 × 23 = 391 right, at about 36 tokens/s using about 1.1 GB of RAM. Numbers measured on our test box, not typical performance.
Try it
ollama run bonsai-2:2bInstall Ollama from ollama.com first; 911 MB download. We ran it with --verbose --think=false added on our CPU-only test box (8 cores, no GPU); speed and memory there are not typical performance.