Yesterday I found this comment on this blog: "As a young researcher trying hard to write nice papers, I don't find many role models."
Forget the introduction, and look at the technical content: what are examples of particularly well-written papers?
I personally like the papers (and lecture notes) by Michel Goemans. I find them clean, rigorous, concise and crisp, elegant in a minimalistic way. For example, his 2006 paper on minimum bounded degree spanning tree (http://math.mit.edu/~goemans/PAPERS/bmst-focs06.pdf). But there may be a cultural bias. Maupassant would be my ideal role model. In his short stories, he always uses precisely the right word to convey the correct meaning. He reputedly spent much time trimming his texts and eliminating all unnecessary adjectives. Why not aim to do the same for technical writing? Write exactly what is necessary for the readers' understanding, no less but no more. Each word should have a purpose, each notation a justification, each sentence a reason for being there. For example each mathematical concept, in my ideal world, is defined exactly twice, once with words and once with symbols, because some readers are more comfortable with words and some with Mathematical notation, and because the two presentations are mutually reinforcing, and help remove possible ambiguities. More than twice is a waste of words, in my ideal world.