{"slug": "apple-foundation-models", "name": "Apple Foundation Models", "category": "ai-models", "category_name": "AI models", "description": "", "recommenders": 3, "commenters": 15, "warners": 4, "score": 0.974, "rising": -0.98, "first": "2026-01-30", "latest": "2026-10-16", "why": "i am more excited about the ondevice foundation model update that is coming", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 3, "neutral": 6, "compare": 2, "warn": 4}, "spark": [0, 0, 0, 0, 0, 0, 1, 2, 0, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "i am more excited about the ondevice foundation model update that is coming", "kind": "", "icon": true, "domain": null, "github_repo": "github.com/johnhenry/apple-foundation-models", "official_url": "https://developer.apple.com/documentation/foundationmodels", "pricing": "", "aliases": ["Apple Foundation Models", "apple-foundation-models"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 775256880, "item_id": 48453555, "url": "https://news.ycombinator.com/item?id=48453555", "by": "mark_l_watson", "date": "2026-06-08", "version": null, "stance": "recommend", "quote": "for small models they are good", "excerpt": "I have been using their on define AFM models for a year - for small models they are good. Their Secure Enclave server bases AFM model is good, but not in the same class as gemini 3.5 flash or deep seek v4 flash.", "context": "has used them for a year", "use_case": "on-device small models", "switched_from": "", "story": "Apple reveals new AI architecture built around Google Gemini models"}, {"id": 775251649, "item_id": 48453228, "url": "https://news.ycombinator.com/item?id=48453228", "by": "franze", "date": "2026-06-08", "version": null, "stance": "recommend", "quote": "i am more excited about the ondevice foundation model update that is coming", "excerpt": "i am more excited about the ondevice foundation model update that is coming https://developer.apple.com/documentation/updates/foundationmodels (not much info yet) but i maintain https://github.com/Arthur-Ficial/apfel so i might be biased", "context": "", "use_case": "on-device AI", "switched_from": "", "story": "Apple Core AI Framework"}, {"id": 768605424, "item_id": 48037839, "url": "https://news.ycombinator.com/item?id=48037839", "by": "syndeo", "date": "2026-05-06", "version": null, "stance": "recommend", "quote": "This is the whole appeal of Apple's Foundation models on iOS too.", "excerpt": "Totally fair. Depends on how it's used, really. If it's to allow the web apps I use to work with more privacy, or to enable smaller/indie players (that can't easily afford to burn a bunch of API tokens for every user) to offer some basic AI-based features in their web apps, then I'm all for that. This is the whole appeal of Apple's Foundation models on iOS too.", "context": "", "use_case": "on-device AI in apps", "switched_from": "", "story": "Google Chrome silently installs a 4 GB AI model on your device without consent"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 801925936, "item_id": 50120371, "url": "https://news.ycombinator.com/item?id=50120371", "by": "cj00", "date": "2026-10-16", "version": null, "stance": "warn", "quote": "I've found some strange guardrails with the Foundation models. Like blocking describing of any photo with a single woman in it.", "excerpt": "I've found some strange guardrails with the Foundation models. Like blocking describing of any photo with a single woman in it. It does this relatively consistently but with a different subset every time. Like out of 10 photos in the set, it's skips 6 then the next run, it skips 3. Have you seen guardrail issues like this? I switched to Gemma 4 instead.", "context": "", "use_case": "photo description", "switched_from": "", "story": "Apple ML Research: Muse Spark Open with MLX on Mac Studio"}, {"id": 799369072, "item_id": 49960567, "url": "https://news.ycombinator.com/item?id=49960567", "by": "redox99", "date": "2026-10-05", "version": null, "stance": "warn", "quote": "the AFM advanced or whatever its called (the 20B local one) sucks", "excerpt": "I love local models, but the AFM advanced or whatever its called (the 20B local one) sucks. And it's not really used in MacOS outside of \"fm\" and very niche stuff. Siri is always cloud even if you're offline.", "context": "", "use_case": "local on-device model", "switched_from": "", "story": "Turn off Apple Intelligence on macOS 27 and get its disk space back"}, {"id": 781703680, "item_id": 48856480, "url": "https://news.ycombinator.com/item?id=48856480", "by": "mathisfun123", "date": "2026-07-09", "version": null, "stance": "warn", "quote": "they just suck", "excerpt": "apple has invested in training models https://machinelearning.apple.com/research/introducing-third-generation-of-apple-foundation-models they just suck", "context": "", "use_case": "", "switched_from": "", "story": "Apple Silicon Exec Explains Mac Mini AI Demand and On-Device Future"}, {"id": 762127249, "item_id": 47632953, "url": "https://news.ycombinator.com/item?id=47632953", "by": "FinnKuhn", "date": "2026-04-03", "version": null, "stance": "warn", "quote": "Ok, then it might just be pretty bad at German Sie/Du.", "excerpt": "Ok, then it might just be pretty bad at German Sie/Du. It manages to translate texts, but is unable to change between Du and Sie. interesting limitation I haven't encountered anywhere before. Even DeepL was able to swap between them. Here is the chat I used to test this again to highlight this limitation: you\u203a write a short email to john asking for a good time to meet ai\u203a Subject: Let's Catch Up Soon! Hi John, I hope", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Apfel \u2013 The free AI already on your Mac"}]}], "who": [{"by": "toppa33", "context": "builds apps", "url": "https://news.ycombinator.com/item?id=50028811", "date": "2026-10-10"}, {"by": "mark_l_watson", "context": "has used them for a year", "url": "https://news.ycombinator.com/item?id=48453555", "date": "2026-06-08"}], "tags": [], "use_cases": ["on-device ai for apps", "photo description", "parsing recipes in apps", "local on-device model", "on-device model cli", "on-device small models"], "compared": [{"slug": "gemma", "name": "Gemma", "co": 3, "switched_from": 0, "switched_to": 1}, {"slug": "qwen", "name": "Qwen", "co": 1, "switched_from": 0, "switched_to": 1}, {"slug": "apfel", "name": "apfel", "co": 3, "switched_from": 0, "switched_to": 0}], "makers": [{"id": 740311568, "item_id": 46269473, "url": "https://news.ycombinator.com/item?id=46269473", "by": "johnhenry", "date": "2025-12-14", "version": null, "stance": "recommend", "quote": "I wrote a small library that exposes apple's LLMs in node", "excerpt": "I wrote a small library that exposes apple's LLMs in node: https://github.com/johnhenry/apple-foundation-models/", "context": "", "use_case": "", "switched_from": "", "story": "Your Mac has a fast, offline LLM"}], "threads": [{"url": "https://news.ycombinator.com/item?id=48450142", "title": "Apple reveals new AI architecture built around Google Gemini models", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=47624645", "title": "Show HN: Apfel \u2013 The free AI already on your Mac", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=50118320", "title": "Apple ML Research: Muse Spark Open with MLX on Mac Studio", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=50080275", "title": "Show HN: Pulse Local, a Mac budgeting app whose AI models run on your machine", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=50016644", "title": "Ask HN: Are you a subscribed LLM or a locally deployed open-source model?", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49957116", "title": "Turn off Apple Intelligence on macOS 27 and get its disk space back", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {"neutral": 1}, "2026-02": {}, "2026-03": {}, "2026-04": {"compare": 1, "warn": 1}, "2026-05": {"recommend": 1}, "2026-06": {"neutral": 3, "recommend": 2}, "2026-07": {"warn": 1}, "2026-08": {}, "2026-09": {"neutral": 1}, "2026-10": {"warn": 2, "neutral": 1, "compare": 1}}, "versions": [], "version_unnamed": 15, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 801925936, "item_id": 50120371, "url": "https://news.ycombinator.com/item?id=50120371", "by": "cj00", "date": "2026-10-16", "version": null, "stance": "warn", "quote": "I've found some strange guardrails with the Foundation models. Like blocking describing of any photo with a single woman in it.", "excerpt": "I've found some strange guardrails with the Foundation models. Like blocking describing of any photo with a single woman in it. It does this relatively consistently but with a different subset every time. Like out of 10 photos in the set, it's skips 6 then the next run, it skips 3. Have you seen guardrail issues like this? I switched to Gemma 4 instead.", "context": "", "use_case": "photo description", "switched_from": "", "story": "Apple ML Research: Muse Spark Open with MLX on Mac Studio"}, {"id": 801296593, "item_id": 50081037, "url": "https://news.ycombinator.com/item?id=50081037", "by": "lathiat", "date": "2026-10-14", "version": null, "stance": "compare", "quote": "Why not leverage Apple Foundational models?", "excerpt": "Why not leverage Apple Foundational models? Downloads will already be on the machine. https://developer.apple.com/documentation/foundationmodels", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Pulse Local, a Mac budgeting app whose AI models run on your machine"}, {"id": 800460977, "item_id": 50028811, "url": "https://news.ycombinator.com/item?id=50028811", "by": "toppa33", "date": "2026-10-10", "version": null, "stance": "neutral", "quote": "I\u2019ve used Apple\u2019s on device foundational models for parsing recipes and other small tasks", "excerpt": "The economics to run a local model don\u2019t stack up for me but I love the idea of running an open model locally for safety & privacy. I use Claude plus I\u2019ve used Apple\u2019s on device foundational models for parsing recipes and other small tasks in the apps I\u2019ve built.", "context": "builds apps", "use_case": "parsing recipes in apps", "switched_from": "", "story": "Ask HN: Are you a subscribed LLM or a locally deployed open-source model?"}, {"id": 799369072, "item_id": 49960567, "url": "https://news.ycombinator.com/item?id=49960567", "by": "redox99", "date": "2026-10-05", "version": null, "stance": "warn", "quote": "the AFM advanced or whatever its called (the 20B local one) sucks", "excerpt": "I love local models, but the AFM advanced or whatever its called (the 20B local one) sucks. And it's not really used in MacOS outside of \"fm\" and very niche stuff. Siri is always cloud even if you're offline.", "context": "", "use_case": "local on-device model", "switched_from": "", "story": "Turn off Apple Intelligence on macOS 27 and get its disk space back"}, {"id": 795194560, "item_id": 49699660, "url": "https://news.ycombinator.com/item?id=49699660", "by": "divbzero", "date": "2026-09-14", "version": null, "stance": "neutral", "quote": "fm = Foundation Models", "excerpt": "fm = Foundation Models https://developer.apple.com/videos/play/wwdc2026/334/", "context": "", "use_case": "on-device model CLI", "switched_from": "", "story": "Apple's Siri AI Can Be Swapped Out for Claude, ChatGPT, Code Shows"}, {"id": 781703680, "item_id": 48856480, "url": "https://news.ycombinator.com/item?id=48856480", "by": "mathisfun123", "date": "2026-07-09", "version": null, "stance": "warn", "quote": "they just suck", "excerpt": "apple has invested in training models https://machinelearning.apple.com/research/introducing-third-generation-of-apple-foundation-models they just suck", "context": "", "use_case": "", "switched_from": "", "story": "Apple Silicon Exec Explains Mac Mini AI Demand and On-Device Future"}, {"id": 776640816, "item_id": 48540051, "url": "https://news.ycombinator.com/item?id=48540051", "by": "scosman", "date": "2026-06-15", "version": null, "stance": "neutral", "quote": "The model the device already has is best", "excerpt": "I think that's what they are trying to avoid. If you need on-device intelligence, their pitch was \"The model the device already has is best\", and if you need something more specific an adapter (aka, a fine-tune/lora) is best. They were wrong when their on-device model was way behind. They still might be right in the long term. While multiple app I use might need Gemma 4 E4B, I use dozens of apps and app devs can choo", "context": "", "use_case": "on-device AI for apps", "switched_from": "", "story": "Apple Foundation Models"}, {"id": 776581971, "item_id": 48536373, "url": "https://news.ycombinator.com/item?id=48536373", "by": "abeymm", "date": "2026-06-14", "version": null, "stance": "neutral", "quote": "Apple Foundation Models", "excerpt": "I created a free golf app with advanced visualization none of the paid apps have. New in the latest release: you can print your personalized Playbook at home (print, cut, fold, staple). I've been playing on a national amateur golf tour for over ten years. Before each event, I'd spend hours on Google Earth \u2014 figuring out where to aim on every hole and which club to hit. It got very cumbersome very soon. I wanted a too", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: What are you working on? (June 2026)"}, {"id": 775256880, "item_id": 48453555, "url": "https://news.ycombinator.com/item?id=48453555", "by": "mark_l_watson", "date": "2026-06-08", "version": null, "stance": "recommend", "quote": "for small models they are good", "excerpt": "I have been using their on define AFM models for a year - for small models they are good. Their Secure Enclave server bases AFM model is good, but not in the same class as gemini 3.5 flash or deep seek v4 flash.", "context": "has used them for a year", "use_case": "on-device small models", "switched_from": "", "story": "Apple reveals new AI architecture built around Google Gemini models"}, {"id": 775251649, "item_id": 48453228, "url": "https://news.ycombinator.com/item?id=48453228", "by": "franze", "date": "2026-06-08", "version": null, "stance": "recommend", "quote": "i am more excited about the ondevice foundation model update that is coming", "excerpt": "i am more excited about the ondevice foundation model update that is coming https://developer.apple.com/documentation/updates/foundationmodels (not much info yet) but i maintain https://github.com/Arthur-Ficial/apfel so i might be biased", "context": "", "use_case": "on-device AI", "switched_from": "", "story": "Apple Core AI Framework"}, {"id": 775227600, "item_id": 48451725, "url": "https://news.ycombinator.com/item?id=48451725", "by": "kube-system", "date": "2026-06-08", "version": null, "stance": "neutral", "quote": "It's a 3B Apple Foundation model.", "excerpt": "> what's actually powering Apple Intelligence now. It's a 3B Apple Foundation model. https://machinelearning.apple.com/research/introducing-apple-foundation-models If you've got a mac, you can use this to play around with it: https://apfel.franzai.com/", "context": "", "use_case": "", "switched_from": "", "story": "Apple reveals new AI architecture built around Google Gemini models"}, {"id": 768605424, "item_id": 48037839, "url": "https://news.ycombinator.com/item?id=48037839", "by": "syndeo", "date": "2026-05-06", "version": null, "stance": "recommend", "quote": "This is the whole appeal of Apple's Foundation models on iOS too.", "excerpt": "Totally fair. Depends on how it's used, really. If it's to allow the web apps I use to work with more privacy, or to enable smaller/indie players (that can't easily afford to burn a bunch of API tokens for every user) to offer some basic AI-based features in their web apps, then I'm all for that. This is the whole appeal of Apple's Foundation models on iOS too.", "context": "", "use_case": "on-device AI in apps", "switched_from": "", "story": "Google Chrome silently installs a 4 GB AI model on your device without consent"}, {"id": 762127249, "item_id": 47632953, "url": "https://news.ycombinator.com/item?id=47632953", "by": "FinnKuhn", "date": "2026-04-03", "version": null, "stance": "warn", "quote": "Ok, then it might just be pretty bad at German Sie/Du.", "excerpt": "Ok, then it might just be pretty bad at German Sie/Du. It manages to translate texts, but is unable to change between Du and Sie. interesting limitation I haven't encountered anywhere before. Even DeepL was able to swap between them. Here is the chat I used to test this again to highlight this limitation: you\u203a write a short email to john asking for a good time to meet ai\u203a Subject: Let's Catch Up Soon! Hi John, I hope", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Apfel \u2013 The free AI already on your Mac"}, {"id": 762061120, "item_id": 47628820, "url": "https://news.ycombinator.com/item?id=47628820", "by": "karimf", "date": "2026-04-03", "version": null, "stance": "compare", "quote": "Apple model is nice since you don't need to download anything else", "excerpt": "The big question is whether Apple can keep shipping new models constantly. AFAIK the current model is on par with with Qwen-3-4B, which is from a year ago [0]. There's a big leap going from last year Qwen-3-4B to Qwen-3.5-4B or to Gemma 4. Apple model is nice since you don't need to download anything else, but I'd rather use the latest model than to use a model from a year ago. https://machinelearning.apple.com/resea", "context": "", "use_case": "On-device LLM", "switched_from": "", "story": "Show HN: Apfel \u2013 The free AI already on your Mac"}, {"id": 749342912, "item_id": 46833932, "url": "https://news.ycombinator.com/item?id=46833932", "by": "wmf", "date": "2026-01-30", "version": null, "stance": "neutral", "quote": "They're also available to other apps", "excerpt": "I think they are also used for translation, summarization, etc. They're also available to other apps: https://developer.apple.com/documentation/FoundationModels", "context": "", "use_case": "on-device AI for apps", "switched_from": "", "story": "The $100B Megadeal Between OpenAI and Nvidia Is on Ice"}]}