MiniCPM5-1B: a 688 MB free model we ran with no GPU
MiniCPM5-1B from OpenBMB has about 1 billion parameters, is free under Apache 2.0 and is built for on-device use; it can think before answering, and on Ollama it's a 688 MB download. Our test on a CPU-only box with 8 cores and no GPU: it correctly said 91 is not prime (7 × 13), but in our recording it missed a Spanish translation request and answered the question instead. Asked fresh, outside the recording, it translated correctly. About 45 tokens/s using about 0.8 GB of RAM; numbers measured on our test box, not typical performance.
Try it
ollama run openbmb/minicpm5Install Ollama from ollama.com first. The model is a 688 MB download. We ran it with --verbose added on our CPU-only test box (8 cores, no GPU); speed and memory there are not typical performance.