# DeepSeekMath-V2

10 Hacker News commenters recommended DeepSeekMath-V2 and 6 criticized it, November 2025 to November 2026. It was recommended in 6 of 13 months.

**Verdict:** Free open model praised for strong math proofs and problem solving; main caveat is it is weak outside math.

## Why people recommend it

- Outperforms larger models on math (3 comments): "DeepSeek R1 and DeepSeekMath were groundbreaking models which were substantially stronger than any models available for distillation at the time they were developed" ([source](https://news.ycombinator.com/item?id=50217516)) "They have had the best math models for about a year most folks just didn't know about it." ([source](https://news.ycombinator.com/item?id=47889681)) "It was a tiny model that outperformed huge then-SOTA LLMs like Claude 3 Opus in math" ([source](https://news.ycombinator.com/item?id=46188534))
- Strong competition math results (2 comments): "Is everyone just glossing over the first place score of 118/120 on the Putnam?" ([source](https://news.ycombinator.com/item?id=46073825)) "the model achieved nearly 99% accuracy" ([source](https://news.ycombinator.com/item?id=46074715))

## What people warn about

- Self-graded results, possible contamination (3 comments): "they are using their own verifier as apanem of experts, with their own model trained specifically to satisfy this verifier" ([source](https://news.ycombinator.com/item?id=46076962)) "They do not mention doing any benchmark decontamination or excluding 2024/2025 competition problems from training." ([source](https://news.ycombinator.com/item?id=46108984)) "it's not a real score; they judged the results themselves and Putnam judges are notoriously tough" ([source](https://news.ycombinator.com/item?id=46076434))
- Narrow, overeager math focus (2 comments): "It's so eager to solve math problems, that it even comes up with random ones if you prompt it with "Hello"." ([source](https://news.ycombinator.com/item?id=46107161)) "This model can’t be used for say questions on biology or history." ([source](https://news.ycombinator.com/item?id=46107750))

## Alternatives hackers recommend

- [Leanstral by Mistral](https://hackerapproved.com/p/leanstral-1-5.md): 15 recommend
- [Harmonic Aristotle](https://hackerapproved.com/p/aristotle.md): 10 recommend

## Cite as

Hacker Approved, "DeepSeekMath-V2", https://hackerapproved.com/p/deepseekmath-v2/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).