EVERY SIGNAL, SORTED.

Denmark, Local AI, Efficient Models, Long Context, Model Architecture and Benchmark Evidence

Denmark's tiny Heimr model stayed fast as its memory grew

Copenhagen lab Sevren says its experimental Heimr 570M model decodes long prompts faster than several small open models. The architecture is interesting. The measurements are promising. Both still need independent reproduction.