# DS4 by antirez 109 Hacker News commenters recommended DS4 by antirez and 20 criticized it, November 2025 to November 2026. It was recommended in 7 of 13 months; owners who said how long report a median of 0.1 years (3 owners). It meets the Hacker Approved badge rule. **Verdict:** Runs DeepSeek V4 Flash on home hardware like Macs, with well-made quants and quality some find close to Claude, though it's slow and its reasoning can derail. ## Why people recommend it - Runs strong model on local hardware (8 comments): "I can run DSv4 Flash on a pair of DGX Sparks and have enough memory left over for 3M tokens of KV cache" ([source](https://news.ycombinator.com/item?id=48848444)) "pretty painless, model runs in about 80GB of RAM and it seemed to be very capable at writing code" ([source](https://news.ycombinator.com/item?id=48142458)) "I've been running Deepseek V4 Flash locally on my Macbook Pro for weeks using Open WebUI + DS4." ([source](https://news.ycombinator.com/item?id=48984004)) - Near frontier-model quality (8 comments): "the value of human ingenuity coupled with a model like DS4.1 is technically enough to get most people to where they want to go" ([source](https://news.ycombinator.com/item?id=50322533)) "It’s shocking how close this feels to claude, obviously it's much slower, but I don’t know that it’s significantly dumber." ([source](https://news.ycombinator.com/item?id=48143174)) "It definitely seems to be the leader on 'general intelligence' on this hardware from my casual usage" ([source](https://news.ycombinator.com/item?id=48863769)) - High-quality low-bit quants (4 comments): "the quality of the 2bit quants antirez has provided have been validated by folks like ggernov of llama.cpp project" ([source](https://news.ycombinator.com/item?id=48684740)) "DS4-flash does, as previously mentioned, handle quantization very well. Even at 2-bit it's still very coherent." ([source](https://news.ycombinator.com/item?id=48848545)) "the ux of ds4 is fantastic too -- it's dead-easy to get a known-good model, great quant" ([source](https://news.ycombinator.com/item?id=48144478)) - Easy setup and good UX (4 comments): "I immediately knew what it was, what it required, and it provided the download links right there." ([source](https://news.ycombinator.com/item?id=49943952)) "pretty painless, model runs in about 80GB of RAM and it seemed to be very capable at writing code" ([source](https://news.ycombinator.com/item?id=48142458)) "It worked, the change I had to do was also minimal for the compilation of the .cu unit." ([source](https://news.ycombinator.com/item?id=49951165)) ## What people warn about - Slow generation and prefill (5 comments): "A little slower than frontier models but my $200/mo weekly usage exhausted with 3 days left on the week" ([source](https://news.ycombinator.com/item?id=48547053)) "DS4 (a 1-bit quantization of DeepSeek V4 Flash) runs at something like 9-13 tokens/second" ([source](https://news.ycombinator.com/item?id=48810949)) "it tops out at something like 10 or 11 with a long time to first response" ([source](https://news.ycombinator.com/item?id=48649558)) - Reasoning loops or goes off-track (6 comments): "it's _good_ at some tasks, but it really just spins its wheels when it hits the limit of what it can reason through" ([source](https://news.ycombinator.com/item?id=49449728)) "it often builds a completely wrong mental model and charges off down the wrong path" ([source](https://news.ycombinator.com/item?id=48854353)) "I wasn't always a fan of what it came up with when asked to plan something." ([source](https://news.ycombinator.com/item?id=49483742)) - Context and memory limits (2 comments): "you'll have a context limit of about 64k, which is also prohibitive for serious work" ([source](https://news.ycombinator.com/item?id=48452384)) "It was like 50GB instead of 5GB for context, and it wasn't fixed for weeks" ([source](https://news.ycombinator.com/item?id=49914489)) ## Alternatives hackers recommend - [llama.cpp](https://hackerapproved.com/p/llama-cpp.md): 471 recommend - [Ollama](https://hackerapproved.com/p/ollama.md): 350 recommend - [LM Studio](https://hackerapproved.com/p/lm-studio.md): 217 recommend - [Unsloth Studio](https://hackerapproved.com/p/unsloth.md): 94 recommend - [oMLX](https://hackerapproved.com/p/omlx.md): 62 recommend ## Cite as Hacker Approved, "DS4 by antirez", https://hackerapproved.com/p/ds4/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).