Chinese AI developer DeepSeek has released its latest model which it said was an “experimental release” that was more efficient to train and better at processing long sequences of text than previous iterations.
DeepSeek releases model it calls ‘intermediate step’ towards ‘next-generation architecture’