# VibeThinker 3B Hacker Approved page: https://hackerapproved.com/p/vibethinker-3b/ 17 people recommended VibeThinker 3B in Hacker News comments, 10 had mixed feelings and 8 criticized it, November 2025 to October 2026. It was recommended in 4 of 12 months, in 7 threads. It does not meet the Hacker Approved badge rule. 12 of the 17 people who recommend it commented in one thread, “VibeThinker: 3B param model that beats Opus 4.5 on reasoning with novel SFT+GRPO”. **Verdict:** Small open model that people recommend for math and logical reasoning, performing well above its size. Common caveats: weak outside its specialty, poor at structured output and tool use. Summary written by Claude from 49 comments. ## Why people recommend it - Strong math and logical reasoning (5 comments in the sample): "basically a match for the larger frontier models -- hundreds of times larger -- in the narrow domain of mathematical and logical reasoning problems" ([source](https://news.ycombinator.com/item?id=49338449)) "performs like models 100x its size on mathematical and logical reasoning tasks" ([source](https://news.ycombinator.com/item?id=49413183)) "It's a densely trained on math whuch let's it punch way higher than it should" ([source](https://news.ycombinator.com/item?id=48611743)) - Strong results for its size (4 comments in the sample): "on well-constrained, verifiable reasoning tasks, first-tier performance is no longer the exclusive domain of ultra-large models" ([source](https://news.ycombinator.com/item?id=48645661)) "Overall it still seems extremely good for its size and I wouldn't expect anything below 30B to behave like that." ([source](https://news.ycombinator.com/item?id=45914512)) "beside a marked tendency towards rambling thinking, it's just plain implausibly good for a 1.5B model" ([source](https://news.ycombinator.com/item?id=45929004)) - Writes working Python code (2 comments in the sample): "It can seemingly reliably write working Python code though" ([source](https://news.ycombinator.com/item?id=48567567)) "It seems weirdly good at some simple python questions." ([source](https://news.ycombinator.com/item?id=45913773)) - Finds valid solutions to problems (1 comment in the sample): "And surprisingly it found a valid solution!" ([source](https://news.ycombinator.com/item?id=48641031)) ## What people warn about - No tool calling capability (3 comments in the sample): "it's trained to generate long CoT to reason through problems (and does it well!) but has ~no tool calling capability" ([source](https://news.ycombinator.com/item?id=48647270)) "I'm not seeing any mention of tools in the paper, much less a bias towards "curiosity" to use those tools" ([source](https://news.ycombinator.com/item?id=48640199)) "This model was not trained on tool-calling or agent-based programming data." ([source](https://news.ycombinator.com/item?id=48650233)) - Rambling thinking, loops, wrong steps (2 comments in the sample): "Some incorrect things and getting into thinking loops, but always coherent" ([source](https://news.ycombinator.com/item?id=45914683)) "it takes so many shortcuts it writes wrong steps in its answers" ([source](https://news.ycombinator.com/item?id=48564032)) - Limited use outside math (2 comments in the sample): "it failed horribly just showing me a rectangle and a black circle" ([source](https://news.ycombinator.com/item?id=48639746)) "the real world use cases for this model seem to be very limited" ([source](https://news.ycombinator.com/item?id=45917295)) - Benchmark gains look like overfitting (1 comment in the sample): "Overfitting, no need to argue about anything I think?" ([source](https://news.ycombinator.com/item?id=48592929)) ## Other reasoning models people recommend - [OpenAI o3](https://hackerapproved.com/p/o3/): 30 recommend, 5 mixed, 11 criticize - [DeepSeek R1](https://hackerapproved.com/p/deepseek-r1/): 18 recommend, 6 mixed, 13 criticize - [OpenAI o1](https://hackerapproved.com/p/openai-o1/): 6 recommend, 6 mixed, 9 criticize ## The badge A product earns the Hacker Approved badge when at least 10 people recommend it, recommenders outnumber critics at least 2 to 1 (mixed feelings count toward neither side), it was recommended in at least 4 of the 12 months and in at least 5 different threads, and no single thread supplies more than half of the people who recommend it. Health and money products are never badged. ## Cite as Hacker Approved, "VibeThinker 3B", https://hackerapproved.com/p/vibethinker-3b/ (17 recommend, 10 mixed, 8 criticize; Hacker News comments, November 2025 to October 2026). Counts updated 2 Nov 2026, with data through October 31, 2026.