Sam Witteveen · 2026-06-19 · notable
Sam Witteveen: VibeThinker 3B — taking on giant models
Sam Witteveen reviews Weibo's VibeThinker-3B, the new 3B-parameter reasoning model that scores 80.2% on LiveCodeBench v6. Witteveen walks through how a 3B open-weights model competes with much larger frontier models on code and math reasoning.
Sam Witteveen's hands-on walkthrough of Weibo's VibeThinker-3B, the small open-weights reasoning model hitting 80.2% on LiveCodeBench v6.
What is it?
Sam Witteveen is a long-time AI YouTuber known for hands-on reviews of new open-weights models. In this video Sam Witteveen breaks down VibeThinker-3B, Weibo's 3-billion-parameter reasoning model that scores 80.2% on LiveCodeBench v6 — a result usually reserved for frontier-scale models.
How does it work?
Sam Witteveen runs through what makes VibeThinker-3B punch above its size: the training recipe, the reasoning behavior, and head-to-head comparisons against larger open and closed models on code and math tasks. The video shows how VibeThinker-3B performs at 3B parameters, which makes the model practical to run on a single consumer GPU.
Why does it matter?
If a 3B model can hit 80.2% on LiveCodeBench v6, the cost-per-token math for hobbyists and small teams shifts. Sam Witteveen's review helps practitioners decide whether VibeThinker-3B is a serious daily driver or a benchmark-overfit curiosity, and shows where the gaps versus larger frontier models actually appear.
Who is it for?
Local-model practitioners, indie devs evaluating small open-weights reasoning models