Spark-X2.6-1.7B: a 1.8 GB free chat model we ran with no GPU
Spark-X2.6-1.7B from SparkLLM/XHToken is an on-device chat model with about 1.7 billion parameters, free under Apache 2.0, with controllable thinking (none / low / high). On Ollama the q4_k_m tag is a 1.8 GB download. Our test on a CPU-only box with 8 cores and no GPU, using --think=low: it correctly said 91 is not prime (7 × 13) and named Canberra as Australia's capital, at about 25–27 tokens/s using about 1.9 GB of RAM; numbers measured on our test box, not typical performance.
Try it
ollama run SparkLLM/Spark-X2.6-1.7B:q4_k_mInstall Ollama from ollama.com first (v0.34.1 or later per the maker). The q4_k_m tag is a 1.8 GB download. Optional: add --think=low --verbose. We ran it on our CPU-only test box (8 cores, no GPU); speed and memory there are not typical performance.