{"slug": "lemonade", "name": "Lemonade", "category": "ai", "category_name": "AI models and assistants", "description": "AMD-backed open-source local server for running LLMs, image, and speech models on PCs.", "recommenders": 5, "commenters": 11, "warners": 0, "score": 6.028, "rising": -1.64, "first": "2025-11-01", "latest": "2026-09-17", "why": "HN commenters recommend Lemonade for running local models on AMD hardware, especially Strix Halo desktops, with ROCm, Vulkan and NPU backends plus text, image and speech generation in one easy install.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 6, "neutral": 5, "compare": 1, "warn": 0}, "spark": [0, 0, 0, 0, 0, 3, 0, 1, 2, 0, 0, 0], "listed": true, "loved": ["AMD/Strix Halo support", "Many backends"], "caveat": "Small default context", "short_quote": "Highly recommend lemonade server if you have a strix halo desktop.", "icon": true, "domain": "lemonade-server.ai", "github_repo": null, "official_url": "https://lemonade-server.ai/", "pricing": "open source", "aliases": ["Lemonade", "lemonade-server", "Lemonade Server"], "verdict": [{"text": "HN commenters recommend Lemonade for running local models on AMD hardware, especially Strix Halo desktops,", "cites": [{"id": 781810976, "item_id": 48863186, "url": "https://news.ycombinator.com/item?id=48863186", "by": "syntaxing", "date": "2026-07-10", "stance": "recommend", "quote": "Highly recommend lemonade server if you have a strix halo desktop.", "excerpt": "Highly recommend lemonade server if you have a strix halo desktop. Been using Qwen3.6-35B @ Q_8 as my main driver and it\u2019s been great with 60 TPS for generation. I occasionally use the 27B @ q6 but only get 20-25 TPS for generation with MTP.", "context": "Strix Halo desktop user", "use_case": "running local LLMs", "switched_from": "", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 765151808, "item_id": 47821988, "url": "https://news.ycombinator.com/item?id=47821988", "by": "seemaze", "date": "2026-04-18", "stance": "recommend", "quote": "Check out the officially supported project Lemonade[0] by AMD.", "excerpt": "Check out the officially supported project Lemonade[0] by AMD. It has gfx1151 specific builds of vLLM, llama.cpp, comfy-ui, and even a PR to merge a Strix Halo port of Apple\u2019s MLX[1] with a quick and easy install. [0] https://www.amd.com/en/developer/resources/technical-articles/2026/lemonade-for-local-ai.html [1] https://github.com/lemonade-sdk/lemonade/issues/1642", "context": "", "use_case": "local AI on Strix Halo", "switched_from": "", "story": "My first impressions on the ROCm and Strix Halo"}, {"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"text": " with ROCm, Vulkan and NPU backends plus text, image and speech generation in one easy install.", "cites": [{"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 765151808, "item_id": 47821988, "url": "https://news.ycombinator.com/item?id=47821988", "by": "seemaze", "date": "2026-04-18", "stance": "recommend", "quote": "Check out the officially supported project Lemonade[0] by AMD.", "excerpt": "Check out the officially supported project Lemonade[0] by AMD. It has gfx1151 specific builds of vLLM, llama.cpp, comfy-ui, and even a PR to merge a Strix Halo port of Apple\u2019s MLX[1] with a quick and easy install. [0] https://www.amd.com/en/developer/resources/technical-articles/2026/lemonade-for-local-ai.html [1] https://github.com/lemonade-sdk/lemonade/issues/1642", "context": "", "use_case": "local AI on Strix Halo", "switched_from": "", "story": "My first impressions on the ROCm and Strix Halo"}, {"id": 761819856, "item_id": 47613741, "url": "https://news.ycombinator.com/item?id=47613741", "by": "freedomben", "date": "2026-04-02", "stance": "recommend", "quote": "Neat, they have rpm, deb, and a companion AppImage desktop app[1]!", "excerpt": "Neat, they have rpm, deb, and a companion AppImage desktop app[1]! Surprised I wasn't aware of this project before. Definitely going to give it a try. [1]: https://github.com/lemonade-sdk/lemonade/releases/tag/v10.0.1", "context": "", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"text": " Caveats include models defaulting to small 4K context windows and lagging behind upstream engine releases,", "cites": [{"id": 781817216, "item_id": 48863576, "url": "https://news.ycombinator.com/item?id=48863576", "by": "data-ottawa", "date": "2026-07-10", "stance": "recommend", "quote": "I used llama swap for a while before leaning into lemonade.", "excerpt": "I second this. I used llama swap for a while before leaning into lemonade. The UI has improved a lot, but be careful as the most of the models default to very small 4K context windows by default. They\u2019re doing some nice things with their Halo models, which load an ensemble of different types of model at the same time. With high vram it\u2019s easy to keep them all in memory, so even though the compute is limited the conte", "context": "", "use_case": "running local LLMs", "switched_from": "llama-swap", "story": "Hands-On with the AMD Ryzen AI Halo"}]}, {"text": " and one user prefers its standalone llama.cpp ROCm build to the combined server.", "cites": [{"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}], "praise": [{"label": "AMD/Strix Halo support", "summary": "Commenters recommend it as the go-to local AI stack for AMD hardware, especially Strix Halo.", "quotes": [{"id": 781810976, "item_id": 48863186, "url": "https://news.ycombinator.com/item?id=48863186", "by": "syntaxing", "date": "2026-07-10", "stance": "recommend", "quote": "Highly recommend lemonade server if you have a strix halo desktop.", "excerpt": "Highly recommend lemonade server if you have a strix halo desktop. Been using Qwen3.6-35B @ Q_8 as my main driver and it\u2019s been great with 60 TPS for generation. I occasionally use the 27B @ q6 but only get 20-25 TPS for generation with MTP.", "context": "Strix Halo desktop user", "use_case": "running local LLMs", "switched_from": "", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 765151808, "item_id": 47821988, "url": "https://news.ycombinator.com/item?id=47821988", "by": "seemaze", "date": "2026-04-18", "stance": "recommend", "quote": "Check out the officially supported project Lemonade[0] by AMD.", "excerpt": "Check out the officially supported project Lemonade[0] by AMD. It has gfx1151 specific builds of vLLM, llama.cpp, comfy-ui, and even a PR to merge a Strix Halo port of Apple\u2019s MLX[1] with a quick and easy install. [0] https://www.amd.com/en/developer/resources/technical-articles/2026/lemonade-for-local-ai.html [1] https://github.com/lemonade-sdk/lemonade/issues/1642", "context": "", "use_case": "local AI on Strix Halo", "switched_from": "", "story": "My first impressions on the ROCm and Strix Halo"}]}, {"label": "Many backends", "summary": "It bundles llama.cpp, vLLM, ComfyUI and other backends covering text, image, TTS and STT across ROCm, Vulkan, CPU, GPU and NPU.", "quotes": [{"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 765151808, "item_id": 47821988, "url": "https://news.ycombinator.com/item?id=47821988", "by": "seemaze", "date": "2026-04-18", "stance": "recommend", "quote": "Check out the officially supported project Lemonade[0] by AMD.", "excerpt": "Check out the officially supported project Lemonade[0] by AMD. It has gfx1151 specific builds of vLLM, llama.cpp, comfy-ui, and even a PR to merge a Strix Halo port of Apple\u2019s MLX[1] with a quick and easy install. [0] https://www.amd.com/en/developer/resources/technical-articles/2026/lemonade-for-local-ai.html [1] https://github.com/lemonade-sdk/lemonade/issues/1642", "context": "", "use_case": "local AI on Strix Halo", "switched_from": "", "story": "My first impressions on the ROCm and Strix Halo"}, {"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"label": "Easy install", "summary": "Users note quick installs, ROCm handled through the app, and rpm, deb and AppImage packages.", "quotes": [{"id": 765151808, "item_id": 47821988, "url": "https://news.ycombinator.com/item?id=47821988", "by": "seemaze", "date": "2026-04-18", "stance": "recommend", "quote": "Check out the officially supported project Lemonade[0] by AMD.", "excerpt": "Check out the officially supported project Lemonade[0] by AMD. It has gfx1151 specific builds of vLLM, llama.cpp, comfy-ui, and even a PR to merge a Strix Halo port of Apple\u2019s MLX[1] with a quick and easy install. [0] https://www.amd.com/en/developer/resources/technical-articles/2026/lemonade-for-local-ai.html [1] https://github.com/lemonade-sdk/lemonade/issues/1642", "context": "", "use_case": "local AI on Strix Halo", "switched_from": "", "story": "My first impressions on the ROCm and Strix Halo"}, {"id": 784341456, "item_id": 49021341, "url": "https://news.ycombinator.com/item?id=49021341", "by": "data-ottawa", "date": "2026-07-23", "stance": "recommend", "quote": "I use Lemonade and have only the ROCm installed through it", "excerpt": "I'm able to use ROCm with MTP just fine (Qwen and Gemma). I use Lemonade and have only the ROCm installed through it (ETA: their llama.cpp version is a fork, not sure what is different vs the main trunk).", "context": "", "use_case": "local LLM inference", "switched_from": "", "story": "New Framework Desktop Option with AMD Ryzen AI Max+ Pro 495 and 192GB Memory"}, {"id": 761819856, "item_id": 47613741, "url": "https://news.ycombinator.com/item?id=47613741", "by": "freedomben", "date": "2026-04-02", "stance": "recommend", "quote": "Neat, they have rpm, deb, and a companion AppImage desktop app[1]!", "excerpt": "Neat, they have rpm, deb, and a companion AppImage desktop app[1]! Surprised I wasn't aware of this project before. Definitely going to give it a try. [1]: https://github.com/lemonade-sdk/lemonade/releases/tag/v10.0.1", "context": "", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"label": "Generation speed", "summary": "Strix Halo users report around 50 to 60 tokens per second on large models.", "quotes": [{"id": 781810976, "item_id": 48863186, "url": "https://news.ycombinator.com/item?id=48863186", "by": "syntaxing", "date": "2026-07-10", "stance": "recommend", "quote": "Highly recommend lemonade server if you have a strix halo desktop.", "excerpt": "Highly recommend lemonade server if you have a strix halo desktop. Been using Qwen3.6-35B @ Q_8 as my main driver and it\u2019s been great with 60 TPS for generation. I occasionally use the 27B @ q6 but only get 20-25 TPS for generation with MTP.", "context": "Strix Halo desktop user", "use_case": "running local LLMs", "switched_from": "", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"label": "Active development", "summary": "Commenters praise a pragmatic development pace and a much-improved UI.", "quotes": [{"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 781817216, "item_id": 48863576, "url": "https://news.ycombinator.com/item?id=48863576", "by": "data-ottawa", "date": "2026-07-10", "stance": "recommend", "quote": "I used llama swap for a while before leaning into lemonade.", "excerpt": "I second this. I used llama swap for a while before leaning into lemonade. The UI has improved a lot, but be careful as the most of the models default to very small 4K context windows by default. They\u2019re doing some nice things with their Halo models, which load an ensemble of different types of model at the same time. With high vram it\u2019s easy to keep them all in memory, so even though the compute is limited the conte", "context": "", "use_case": "running local LLMs", "switched_from": "llama-swap", "story": "Hands-On with the AMD Ryzen AI Halo"}]}, {"label": "Compatible APIs", "summary": "OpenAI- and Ollama-compatible endpoints let it work with VSCode Copilot and Open WebUI.", "quotes": [{"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}], "caveats": [{"label": "Small default context", "summary": "Most models default to very small 4K context windows.", "quotes": [{"id": 781817216, "item_id": 48863576, "url": "https://news.ycombinator.com/item?id=48863576", "by": "data-ottawa", "date": "2026-07-10", "stance": "recommend", "quote": "I used llama swap for a while before leaning into lemonade.", "excerpt": "I second this. I used llama swap for a while before leaning into lemonade. The UI has improved a lot, but be careful as the most of the models default to very small 4K context windows by default. They\u2019re doing some nice things with their Halo models, which load an ensemble of different types of model at the same time. With high vram it\u2019s easy to keep them all in memory, so even though the compute is limited the conte", "context": "", "use_case": "running local LLMs", "switched_from": "llama-swap", "story": "Hands-On with the AMD Ryzen AI Halo"}]}, {"label": "Upstream lag", "summary": "Bundled engines lag upstream releases, its llama.cpp is a fork, and vLLM support is experimental.", "quotes": [{"id": 781817216, "item_id": 48863576, "url": "https://news.ycombinator.com/item?id=48863576", "by": "data-ottawa", "date": "2026-07-10", "stance": "recommend", "quote": "I used llama swap for a while before leaning into lemonade.", "excerpt": "I second this. I used llama swap for a while before leaning into lemonade. The UI has improved a lot, but be careful as the most of the models default to very small 4K context windows by default. They\u2019re doing some nice things with their Halo models, which load an ensemble of different types of model at the same time. With high vram it\u2019s easy to keep them all in memory, so even though the compute is limited the conte", "context": "", "use_case": "running local LLMs", "switched_from": "llama-swap", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 784341456, "item_id": 49021341, "url": "https://news.ycombinator.com/item?id=49021341", "by": "data-ottawa", "date": "2026-07-23", "stance": "recommend", "quote": "I use Lemonade and have only the ROCm installed through it", "excerpt": "I'm able to use ROCm with MTP just fine (Qwen and Gemma). I use Lemonade and have only the ROCm installed through it (ETA: their llama.cpp version is a fork, not sure what is different vs the main trunk).", "context": "", "use_case": "local LLM inference", "switched_from": "", "story": "New Framework Desktop Option with AMD Ryzen AI Max+ Pro 495 and 192GB Memory"}]}, {"label": "Combined server", "summary": "One user dislikes the combined server and uses only its llama.cpp ROCm build instead.", "quotes": [{"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"label": "Limited NPU benefit", "summary": "The NPU targets small always-on models, so it is not a big win for standard chatbot use.", "quotes": [{"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}, {"label": "Unclear positioning", "summary": "One commenter could not tell what it is from the website or whether it supports Macs.", "quotes": [{"id": 761819904, "item_id": 47613744, "url": "https://news.ycombinator.com/item?id=47613744", "by": "JSR_FDED", "date": "2026-04-02", "stance": "compare", "quote": "An alternative to LM Studio? Does it support MLX or metal on Macs?", "excerpt": "I\u2019ve read the website and the news announcement, and I still don\u2019t understand what it is. An alternative to LM Studio? Does it support MLX or metal on Macs? I\u2019m assuming it will optimize things for AMD, but are you at a disadvantage using other GPUs?", "context": "", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}]}], "who": [{"by": "selicos", "context": "infra engineer, local LLM tinkerer", "url": "https://news.ycombinator.com/item?id=49744479", "date": "2026-09-17"}, {"by": "SomeHacker44", "context": "runs 9950X with 192G RAM on Arch", "url": "https://news.ycombinator.com/item?id=48967915", "date": "2026-07-19"}, {"by": "syntaxing", "context": "Strix Halo desktop user", "url": "https://news.ycombinator.com/item?id=48863186", "date": "2026-07-10"}, {"by": "data-ottawa", "context": "Framework Desktop owner", "url": "https://news.ycombinator.com/item?id=48734580", "date": "2026-06-30"}, {"by": "zdragnar", "context": "owns an AMD Strix Halo desktop", "url": "https://news.ycombinator.com/item?id=48723407", "date": "2026-06-29"}, {"by": "rpdillon", "context": "runs a Strix Halo box", "url": "https://news.ycombinator.com/item?id=47614043", "date": "2026-04-02"}, {"by": "dennemark", "context": "uses AMD Strix Halo", "url": "https://news.ycombinator.com/item?id=47613997", "date": "2026-04-02"}], "tags": ["for AMD Strix Halo", "for Linux users", "OpenAI-compatible API", "for local coding assistants"], "use_cases": ["running local llms", "local llm serving", "local llm server", "local llm inference", "running local models", "local ai on strix halo"], "compared": [{"slug": "llama-swap", "name": "llama-swap", "co": 2, "switched_from": 2, "switched_to": 0}, {"slug": "gemma", "name": "Gemma", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "lm-studio", "name": "LM Studio", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "llama-cpp", "name": "llama.cpp", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "rocm", "name": "ROCm", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=47612724", "title": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU", "mentions": 4}, {"url": "https://news.ycombinator.com/item?id=48861833", "title": "Hands-On with the AMD Ryzen AI Halo", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49740105", "title": "Show HN: Share your AI Setup, Learn from others", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49019694", "title": "New Framework Desktop Option with AMD Ryzen AI Max+ Pro 495 and 192GB Memory", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48966190", "title": "Qwen 3.8 Max Preview", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48732679", "title": "DGX Spark vs. Mac Studio and Halo", "mentions": 1}], "sentiment_by_month": {"2025-11": {"neutral": 1}, "2025-12": {}, "2026-01": {}, "2026-02": {}, "2026-03": {}, "2026-04": {"recommend": 3, "compare": 1, "neutral": 1}, "2026-05": {}, "2026-06": {"neutral": 1, "recommend": 1}, "2026-07": {"recommend": 2, "neutral": 1}, "2026-08": {}, "2026-09": {"neutral": 1}, "2026-10": {}}, "quotes": [{"id": 795911665, "item_id": 49744479, "url": "https://news.ycombinator.com/item?id=49744479", "by": "selicos", "date": "2026-09-17", "stance": "neutral", "quote": "Using: LM Studio (primary), Lemonade", "excerpt": "So I have to connect over MCP or log into github through the page to share anything? How about being able to post a markdown summary of my setup instead of plugging my credentials and tools into a third party site? Strix HALO 128gb - Framework mainboard in a custom SFF PC. Just moved from Ubuntu to Fedora 44. Using: LM Studio (primary), Lemonade, not yet got into vLLM and llama.cpp directyly after moving to Fedora. R", "context": "infra engineer, local LLM tinkerer", "use_case": "running local LLMs", "switched_from": "", "story": "Show HN: Share your AI Setup, Learn from others"}, {"id": 784341456, "item_id": 49021341, "url": "https://news.ycombinator.com/item?id=49021341", "by": "data-ottawa", "date": "2026-07-23", "stance": "recommend", "quote": "I use Lemonade and have only the ROCm installed through it", "excerpt": "I'm able to use ROCm with MTP just fine (Qwen and Gemma). I use Lemonade and have only the ROCm installed through it (ETA: their llama.cpp version is a fork, not sure what is different vs the main trunk).", "context": "", "use_case": "local LLM inference", "switched_from": "", "story": "New Framework Desktop Option with AMD Ryzen AI Max+ Pro 495 and 192GB Memory"}, {"id": 783486640, "item_id": 48967915, "url": "https://news.ycombinator.com/item?id=48967915", "by": "SomeHacker44", "date": "2026-07-19", "stance": "neutral", "quote": "I used lemonade with Vulkan and the UD-Q8_x_x model from HF.", "excerpt": "Would you mind sharing more please? I literally just finished the same set up, with a 9950X CPU and 192G RAM at 4,800 MT/s. I used lemonade with Vulkan and the UD-Q8_x_x model from HF. 256k context. I have about 8G VRAM free, and use the iGPU for my desktop/monitor on Arch. What options do you give llama-cpp or whatever you run please? What other models have you found fit nicely in the 2xR9700? Thanks!!", "context": "runs 9950X with 192G RAM on Arch", "use_case": "running local models", "switched_from": "", "story": "Qwen 3.8 Max Preview"}, {"id": 781817216, "item_id": 48863576, "url": "https://news.ycombinator.com/item?id=48863576", "by": "data-ottawa", "date": "2026-07-10", "stance": "recommend", "quote": "I used llama swap for a while before leaning into lemonade.", "excerpt": "I second this. I used llama swap for a while before leaning into lemonade. The UI has improved a lot, but be careful as the most of the models default to very small 4K context windows by default. They\u2019re doing some nice things with their Halo models, which load an ensemble of different types of model at the same time. With high vram it\u2019s easy to keep them all in memory, so even though the compute is limited the conte", "context": "", "use_case": "running local LLMs", "switched_from": "llama-swap", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 781810976, "item_id": 48863186, "url": "https://news.ycombinator.com/item?id=48863186", "by": "syntaxing", "date": "2026-07-10", "stance": "recommend", "quote": "Highly recommend lemonade server if you have a strix halo desktop.", "excerpt": "Highly recommend lemonade server if you have a strix halo desktop. Been using Qwen3.6-35B @ Q_8 as my main driver and it\u2019s been great with 60 TPS for generation. I occasionally use the 27B @ q6 but only get 20-25 TPS for generation with MTP.", "context": "Strix Halo desktop user", "use_case": "running local LLMs", "switched_from": "", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 779753280, "item_id": 48734580, "url": "https://news.ycombinator.com/item?id=48734580", "by": "data-ottawa", "date": "2026-06-30", "stance": "recommend", "quote": "lemonade is just an easier to manage system", "excerpt": "I have the Framework Desktop with 395+ 128gb RAM Today I am pretty happy with it. LLMs are finally good enough (fast enough with MTP+MoE, but also just much better in capability) that I can fit local ones into real tasks, and I've used image generation with invokeAI to do some genuinely useful things like rendering concepts for a renovation. I mostly use lemonade-server and invokeAI for my workloads, previously I use", "context": "Framework Desktop owner", "use_case": "local LLM serving", "switched_from": "llama-swap", "story": "DGX Spark vs. Mac Studio and Halo"}, {"id": 779574513, "item_id": 48723407, "url": "https://news.ycombinator.com/item?id=48723407", "by": "zdragnar", "date": "2026-06-29", "stance": "neutral", "quote": "I haven't tried it with https://lemonade-server.ai/ yet but I just might give it a shot.", "excerpt": "I got mine at the same price point, and I've been pretty pleased with it. Tailscale lets me use it from my ultrabook / lightweight laptop, no burning lap or crazy fan noises. Desktops with the amd ai+ 395 are still fairly affordable for what they can do. I haven't tried it with https://lemonade-server.ai/ yet but I just might give it a shot.", "context": "owns an AMD Strix Halo desktop", "use_case": "local LLM serving", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 765151808, "item_id": 47821988, "url": "https://news.ycombinator.com/item?id=47821988", "by": "seemaze", "date": "2026-04-18", "stance": "recommend", "quote": "Check out the officially supported project Lemonade[0] by AMD.", "excerpt": "Check out the officially supported project Lemonade[0] by AMD. It has gfx1151 specific builds of vLLM, llama.cpp, comfy-ui, and even a PR to merge a Strix Halo port of Apple\u2019s MLX[1] with a quick and easy install. [0] https://www.amd.com/en/developer/resources/technical-articles/2026/lemonade-for-local-ai.html [1] https://github.com/lemonade-sdk/lemonade/issues/1642", "context": "", "use_case": "local AI on Strix Halo", "switched_from": "", "story": "My first impressions on the ROCm and Strix Halo"}, {"id": 761824688, "item_id": 47614043, "url": "https://news.ycombinator.com/item?id=47614043", "by": "rpdillon", "date": "2026-04-02", "stance": "neutral", "quote": "Been running lemonade for some time on my Strix Halo box.", "excerpt": "Been running lemonade for some time on my Strix Halo box. It dispatches out to other backends that they include, like diffusion and llama. I actually don't like their combined server, and what I use instead is their llama CPP build for ROCm. https://github.com/lemonade-sdk/llamacpp-rocm But I'm not doing anything with images or audio. I get about 50 tokens a second with GPT OSS 120B. As others have pointed out, the N", "context": "runs a Strix Halo box", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 761823952, "item_id": 47613997, "url": "https://news.ycombinator.com/item?id=47613997", "by": "dennemark", "date": "2026-04-02", "stance": "recommend", "quote": "Really recommend this for AMD hardware!", "excerpt": "I have been using lemonade for nearly a year already. On Strix Halo I am using nothing else - although kyuz0's toolboxes are also nice (https://kyuz0.github.io/amd-strix-halo-toolboxes/) Nowadays you get TTS, STT, text & image generation and image editing should also be possible. Besides being able to run via rocm, vulkan or on CPU, GPU and NPU. Quite a lot of options. They have a quite good and pragmatic pace in dev", "context": "uses AMD Strix Halo", "use_case": "local LLM server on Strix Halo", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 761819904, "item_id": 47613744, "url": "https://news.ycombinator.com/item?id=47613744", "by": "JSR_FDED", "date": "2026-04-02", "stance": "compare", "quote": "An alternative to LM Studio? Does it support MLX or metal on Macs?", "excerpt": "I\u2019ve read the website and the news announcement, and I still don\u2019t understand what it is. An alternative to LM Studio? Does it support MLX or metal on Macs? I\u2019m assuming it will optimize things for AMD, but are you at a disadvantage using other GPUs?", "context": "", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 761819856, "item_id": 47613741, "url": "https://news.ycombinator.com/item?id=47613741", "by": "freedomben", "date": "2026-04-02", "stance": "recommend", "quote": "Neat, they have rpm, deb, and a companion AppImage desktop app[1]!", "excerpt": "Neat, they have rpm, deb, and a companion AppImage desktop app[1]! Surprised I wasn't aware of this project before. Definitely going to give it a try. [1]: https://github.com/lemonade-sdk/lemonade/releases/tag/v10.0.1", "context": "", "use_case": "local LLM server", "switched_from": "", "story": "Lemonade by AMD: a fast and open source local LLM server using GPU and NPU"}, {"id": 732508578, "item_id": 45781786, "url": "https://news.ycombinator.com/item?id=45781786", "by": "Balinares", "date": "2025-11-01", "stance": "neutral", "quote": "", "excerpt": "Does Qwen3 Coder do a good job invoking its tools as appropriate for you? Under continue.dev at least, I've found I need to remind it constantly.", "context": "", "use_case": "local LLM serving", "switched_from": "", "story": "Ask HN: Who uses open LLMs and coding assistants locally? Share setup and laptop"}]}