# DeepSeek V4 250 Hacker News commenters recommended DeepSeek V4 and 74 criticized it, November 2025 to November 2026. It was recommended in 9 of 13 months; owners who said how long report a median of 0.08 years (5 owners). It meets the Hacker Approved badge rule. **Verdict:** A very cheap open-weights model with efficient long context that many find close to frontier models, though critics say it lags Claude and GPT and hallucinates often. ## Why people recommend it - Far cheaper than Western models (8 comments): "Based solely on quality and price, OpenAI, Anthropic, and other western models just can't compete with the new generation of Chinese open models." ([source](https://news.ycombinator.com/item?id=48492839)) "I'm also experimenting with DeepSeek 4 and it seems on par. Its API price is much cheaper and it's top up, not a subscription." ([source](https://news.ycombinator.com/item?id=48808590)) "Given deepseek-v4 is also the cheapest model among those five, I would say—based on these data—it's the winner." ([source](https://news.ycombinator.com/item?id=48642815)) - Near frontier-level quality (7 comments): "I've been fooling around with DeepSeek 4 agentically. It's probably not as good as Anthropic offerings, but even those seem to be roiled in politics and strife and DeepSeek 4 is very good IMHO." ([source](https://news.ycombinator.com/item?id=48668160)) "To me it feels like GLM 5.1, Kimi 2.6, DeepSeek 4 are all competitive both with each other and with the American models." ([source](https://news.ycombinator.com/item?id=47993700)) "It is more than good enough and has effectively caught up with Opus 4.6 and GPT 5.4 according to the benchmarks." ([source](https://news.ycombinator.com/item?id=47885430)) - Open weights you can self-host (5 comments): "The models are open source open weights, which mean you can inspect what the models do and you you can choose US or EU infra providers" ([source](https://news.ycombinator.com/item?id=48088977)) "open weights is really good, especially when base models (not just the instruction tuned) weights are released like here" ([source](https://news.ycombinator.com/item?id=47887231)) "without having to worry about token cost or about subscription limits or about any future degradation in performance" ([source](https://news.ycombinator.com/item?id=47890519)) - Memory-light KV cache, long context (6 comments): "DeepSeek has a phenomenal cacheing architecture that specifically can work with less-than-ideal clients like this and get cache hits anyway." ([source](https://news.ycombinator.com/item?id=48980284)) "DeepSeek V4 series, which uses an even more advanced attention mechanism and can save a lot of memory capacity for KV cache" ([source](https://news.ycombinator.com/item?id=48640496)) "DeepSeekV4 just came up with a new way to have super long context with KV cache an order of magnitude smaller than before." ([source](https://news.ycombinator.com/item?id=47975388)) ## What people warn about - Lags top Claude and GPT models (4 comments): "I think it's great, but coming from Claude Code it did feel like going back in time by ~6 months in model capabilities." ([source](https://news.ycombinator.com/item?id=48336610)) "OOS models are vool, cheap and promising for a future -- but why are we pretending they are better than they are?" ([source](https://news.ycombinator.com/item?id=47984698)) "It’s also not nearly as good as 5.5. I’ve had 3 different tasks just this week that deepseek has failed at" ([source](https://news.ycombinator.com/item?id=47975289)) - High hallucination rate (2 comments): "Deepseek is one of the worst in terms of hallucination rate according to artificial analysis' benchmark" ([source](https://news.ycombinator.com/item?id=48967996)) "DeepSeek V4 hallucinates like crazy and often forgets explicitly mentioned parts of the context." ([source](https://news.ycombinator.com/item?id=48966513)) - Censorship and Chinese data concerns (4 comments): "Other sensitive questions (eg: Tiananmen Square) it clams up unless you ask the question a specific way" ([source](https://news.ycombinator.com/item?id=49025631)) "they very obviously waited long enough for the GAB to add several dozen xz-level easter eggs to it" ([source](https://news.ycombinator.com/item?id=47886111)) "data privacy, or how this AI may someday be used to undermine US national or economic security" ([source](https://news.ycombinator.com/item?id=47893333)) - Inconsistent language and response format (2 comments): "It will promise to keep it in English from here on out, then switch back to Chinese after two or three exchanges." ([source](https://news.ycombinator.com/item?id=49627979)) "both would do things like change the response format from one message to another in the same chat" ([source](https://news.ycombinator.com/item?id=49201471)) ## Alternatives hackers recommend - [Claude Opus](https://hackerapproved.com/p/claude-opus.md): 3331 recommend - [OpenAI GPT](https://hackerapproved.com/p/gpt-5-5.md): 1205 recommend - [Claude Sonnet](https://hackerapproved.com/p/claude-sonnet.md): 854 recommend - [DeepSeek V4 Flash](https://hackerapproved.com/p/deepseek-v4-flash.md): 828 recommend - [Fable](https://hackerapproved.com/p/fable.md): 823 recommend ## Where to get it - [Visit the official site](https://www.deepseek.com/) ## Cite as Hacker Approved, "DeepSeek V4", https://hackerapproved.com/p/deepseek-v4/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).