WHY WEIBO'S TINY VIBETHINKER-3B HAS THE AI WORLD ARGUING OVER BENCHMARKS AGAIN

Summary from the TLDR AI newsletter

The 3B parameter model put up coding benchmark scores in the same league as Claude Opus 4.5.