B, a 3-billion-parameter AI model, is challenging OpenAI, Google and DeepSeek on math and coding benchmarks while reigniting the debate over AI scaling, benchmark gaming and small-model reasoning.

Concept In a tree, the unique path between two nodes u and v goes up from u to their lowest common ancestor (LCA) and then down to v. If we know the depth of each node and can jump upwards in powers ...

VibeThinker-3B, released by WeiboAI (China's Weibo), claims to have '76.4 on IMO-AnswerBench and a 96.1% pass rate on LeetCode weekly contests.' If this is true for a lightweight 3B model, it is quite ...

There was an error while loading. Please reload this page.