# oMLX 62 Hacker News commenters recommended oMLX and 11 criticized it, November 2025 to November 2026. It was recommended in 8 of 13 months. It meets the Hacker Approved badge rule. **Verdict:** Recommended for fast local models on Apple Silicon, with KV caching that keeps long agent sessions quick. Critics find it slower than llama.cpp in some tests and sometimes unstable. ## Why people recommend it - KV caching keeps long chats fast (11 comments): "much more performant and efficient than LM Studio. It has block-level KV context caching that makes long chats and agentic/tool calling scenarios MUCH faster" ([source](https://news.ycombinator.com/item?id=47870291)) "What really has been a game changer for my workflow is using https://omlx.ai/ that has SSD KV cold caching." ([source](https://news.ycombinator.com/item?id=47583733)) "The good cacheing layer in oMLX will help things feel faster for some classes of tasks in my experience." ([source](https://news.ycombinator.com/item?id=49300506)) - Fast local inference on Apple Silicon (4 comments): "It truly is the SOTA for local inference on mac. Even when there are regressions the dev(s) are insanely responsive." ([source](https://news.ycombinator.com/item?id=48508270)) "offer some amazing performance optimizations for Apple Silicon that utilize UMA and NVME caching efficiently" ([source](https://news.ycombinator.com/item?id=49247723)) "I switched to oMLX I get ~75 tok/s on Qwen3.6-35B-A3B-4bit on a 48GB M5 Pro" ([source](https://news.ycombinator.com/item?id=49329095)) - Easier setup and friendly UI (4 comments): "It's a lot smoother setup than Ollama and has a built in model sizer and all that in a nice ui." ([source](https://news.ycombinator.com/item?id=49653001)) "makes running the mlx inference server quite easy for those interested in UI-based hosting" ([source](https://news.ycombinator.com/item?id=48507775)) "gives you intelligent local hardware based model download recommendations" ([source](https://news.ycombinator.com/item?id=48509077)) - OpenAI-compatible endpoints work with tools (3 comments): "I've been able to run local, quantized LLMs like Qwen 3.6 using oMLX + Opencode pretty comfortably and productively." ([source](https://news.ycombinator.com/item?id=48662900)) "They expose „open ai compatible“ endpoints, so you can configure almost any harness to use them." ([source](https://news.ycombinator.com/item?id=49153622)) "Omlx lets you run Claude code locally though as long as you bootstrap it with the right model" ([source](https://news.ycombinator.com/item?id=48587359)) ## What people warn about - Slower than llama.cpp or LM Studio (3 comments): "MLX in the forms I have tried (LM Studio, oMLX) with the models I have tried (Gemma and Qwen) have both been apparently slower, yes." ([source](https://news.ycombinator.com/item?id=48518405)) "I have not, on a 64GB M1 Max, seen any advantage from oMLX specifically or MLX generally over GGUF with llama.cpp." ([source](https://news.ycombinator.com/item?id=48509499)) "oMLX's token generation is 3–7x slower than llama.cpp's Metal backend." ([source](https://news.ycombinator.com/item?id=47668849)) - Crashes and stability problems (3 comments): "it horribly kernel panics when it is running for too long" ([source](https://news.ycombinator.com/item?id=48015401)) "I returned back to Ollama because of stability issues." ([source](https://news.ycombinator.com/item?id=49653361)) "keeps driving me up a wall with odd papercuts, bugs, and silent failures" ([source](https://news.ycombinator.com/item?id=48985482)) - Bugs with some models and caches (3 comments): "generation is supported but not all special tokens are handled" ([source](https://news.ycombinator.com/item?id=47627460)) "the oMLX performance with the MLX community MTP models is much worse" ([source](https://news.ycombinator.com/item?id=48441450)) "Qwen3.8 27B seems to have weird memory and caching issues on oMLX" ([source](https://news.ycombinator.com/item?id=49528749)) ## 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 - [DS4 by antirez](https://hackerapproved.com/p/ds4.md): 109 recommend - [Unsloth Studio](https://hackerapproved.com/p/unsloth.md): 94 recommend ## Cite as Hacker Approved, "oMLX", https://hackerapproved.com/p/omlx/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).