A compact 7.85M-parameter causal language model with a novel Channel-Mixing Attention (CMA) architecture. Enter a prompt and the model will generate a continuation.