Understanding How DeepSeek's Flagship Open-Weight Models Evolved

Note that this does assume some prior transformer architecture knowledge, but if you know how attention works then you should at least be able to get the overall idea. Comments