# DeepSeek V3

59 Hacker News commenters recommended DeepSeek V3 and 17 criticized it, November 2025 to November 2026. It was recommended in 9 of 13 months. It meets the Hacker Approved badge rule.

**Verdict:** Recommended as a cheap, open-weight model that gets close to closed frontier models on many tasks. Common caveat: slow with large contexts.

## Why people recommend it

- Much cheaper than closed models (6 comments): "DeepSeek beats all others (single-shot), and it is ~50% cheaper than the cost of local electricity only." ([source](https://news.ycombinator.com/item?id=47537903)) "I can fire up DeepSeek 3.2 or GLM 4.6 for a fraction of the cost and get almost as good as results" ([source](https://news.ycombinator.com/item?id=46240131)) "Models like dsv3.2 are really really cheap to serve, for what they provide" ([source](https://news.ycombinator.com/item?id=46337509))
- Open weights you can build on (4 comments): "deepseek has thus far always delivered. And not just delivered, but under open license too" ([source](https://news.ycombinator.com/item?id=46114980)) "they are a good bet against monopolisation by proprietary vendors" ([source](https://news.ycombinator.com/item?id=46111595)) "it's open source, you can download the weights and build on those" ([source](https://news.ycombinator.com/item?id=46115049))
- Close to top models in quality (6 comments): "The current SotA for open models (dsv3, glm, minimax, devstral, etc) are at or above the mini versions of top labs" ([source](https://news.ycombinator.com/item?id=46734923)) "I can fire up DeepSeek 3.2 or GLM 4.6 for a fraction of the cost and get almost as good as results" ([source](https://news.ycombinator.com/item?id=46240131)) "Deepseek 3.2 is a big step up over 3.1, if benchmarks are to be believed" ([source](https://news.ycombinator.com/item?id=46125893))
- Reliable, transparent releases from DeepSeek (5 comments): "DeepSeek has good cred so far on delivering roughly what their PR says they are delivering" ([source](https://news.ycombinator.com/item?id=46113616)) "deepseek has thus far always delivered. And not just delivered, but under open license too" ([source](https://news.ycombinator.com/item?id=46114980)) "winning on cost-effectiveness, and continuing to publicly share their improvements" ([source](https://news.ycombinator.com/item?id=46111149))

## What people warn about

- Trails closed models on accuracy (4 comments): "even if you use SOTA open source models like Kimi K2.5 or Deepseek V3.2 they simply aren't as capable" ([source](https://news.ycombinator.com/item?id=46917848)) "They're not that close (on things like LMArena) and being cheaper is pretty meaningless" ([source](https://news.ycombinator.com/item?id=46112678)) "yet in my testing even sonnet is miles ahead both in terms of speed and accuracy" ([source](https://news.ycombinator.com/item?id=46222124))
- Slow token generation (3 comments): "6 tok/sec isn't acceptable for anything else but async/batched stuff" ([source](https://news.ycombinator.com/item?id=46120155)) "6t/s will have you pulling your hair out with any deepseek model." ([source](https://news.ycombinator.com/item?id=46122060)) "token generation is pretty good. Just large contexts are slow" ([source](https://news.ycombinator.com/item?id=46112394))
- High token use raises costs (2 comments): "the actual amount of tokens required to produce results is high so total costs were actually higher" ([source](https://news.ycombinator.com/item?id=47148539)) "DS-Speciale is 1st or 2nd in accuracy in all tests, but has much higher token output" ([source](https://news.ycombinator.com/item?id=46107926))

## Alternatives hackers recommend

- [Qwen 3.8 27B](https://hackerapproved.com/p/qwen-3-8-27b.md): 1173 recommend
- [Z.ai GLM](https://hackerapproved.com/p/glm-5-2.md): 1043 recommend
- [Kimi](https://hackerapproved.com/p/kimi-k3.md): 931 recommend
- [Gemma 4](https://hackerapproved.com/p/gemma-4.md): 713 recommend
- [Qwen](https://hackerapproved.com/p/qwen.md): 474 recommend

## Cite as

Hacker Approved, "DeepSeek V3", https://hackerapproved.com/p/deepseek-v3-2/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).