{"slug": "needle", "name": "Needle", "category": "ai-models", "category_name": "AI models", "description": "", "recommenders": 6, "commenters": 7, "warners": 0, "score": 5.936, "rising": 3.35, "first": "2026-05-12", "latest": "2026-08-25", "why": "I used Needle as a router in https://rcarmo.github.io/projects/memento/", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 6, "neutral": 1, "compare": 0, "warn": 0}, "spark": [0, 0, 0, 0, 0, 0, 1, 0, 1, 4, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "I used Needle as a router in https://rcarmo.github.io/projects/memento/", "kind": "", "icon": false, "domain": "cactuscompute.com", "github_repo": null, "official_url": "", "pricing": "", "aliases": ["Needle"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 790715648, "item_id": 49419728, "url": "https://news.ycombinator.com/item?id=49419728", "by": "stymaar", "date": "2026-08-24", "version": null, "stance": "recommend", "quote": "I'm a big fan of Cactus's work on the needle family of Simple Attention Network", "excerpt": "True. I'm a big fan of Cactus's work on the needle family of Simple Attention Network: https://github.com/cactus-compute/needle", "context": "", "use_case": "", "switched_from": "", "story": "I were 17, I'd learn how to build LLMs from scratch"}, {"id": 789169552, "item_id": 49323097, "url": "https://news.ycombinator.com/item?id=49323097", "by": "msdz", "date": "2026-08-16", "version": null, "stance": "recommend", "quote": "Cactus have come up with Needle [0][1], which is their tool-calling focused 14 MB model", "excerpt": "Great article. > There's a version of this future where the model card stops listing a knowledge cutoff at all, because what's left in the weights goes stale on a scale of years instead of weeks. Future? Even just recently I\u2019ve read of two approaches to this problem: Cactus have come up with Needle [0][1], which is their tool-calling focused 14 MB model (still an LLM!) \u2013 no world knowledge engrained. And instead of s", "context": "", "use_case": "", "switched_from": "", "story": "Models Are Getting Dumber on Purpose"}, {"id": 788118928, "item_id": 49257433, "url": "https://news.ycombinator.com/item?id=49257433", "by": "a022311", "date": "2026-08-11", "version": null, "stance": "recommend", "quote": "Needle and similar models solve this problem nicely", "excerpt": "Looks incredibly cool, although it ran at ~4 tok/s on my browser (I assume because I've disabled V8). I believe your numbers though. I always wanted to try using something like this to create an AI assistant for my phone. If you try to create something without AI, you'll quickly run into the problem of using NLP to map input to the equivalent of AI tool calls. On the other hand, shelling out to a full-blown LLM for e", "context": "", "use_case": "cheap local tool-call mapping", "switched_from": "", "story": "Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots"}, {"id": 788067696, "item_id": 49254231, "url": "https://news.ycombinator.com/item?id=49254231", "by": "rcarmo", "date": "2026-08-10", "version": null, "stance": "recommend", "quote": "I used Needle as a router in https://rcarmo.github.io/projects/memento/", "excerpt": "Nice. I used Needle as a router in https://rcarmo.github.io/projects/memento/, need to take a look at this", "context": "", "use_case": "LLM router", "switched_from": "", "story": "Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots"}, {"id": 785073568, "item_id": 49067098, "url": "https://news.ycombinator.com/item?id=49067098", "by": "docheinestages", "date": "2026-07-27", "version": null, "stance": "recommend", "quote": "Think of Cactus Compute's Needle", "excerpt": "Labs are focusing on creating models, small or large, that perform well on various benchmarks, including general knowledge, domain-specific expertise, and agentic capabilities. Asking for such a model while wanting to be small and fast would align with what you're describing, which I believe is different from what I'm pointing to. The model I'm describing sacrifices domain knowledge and expertise for agentic reasonin", "context": "", "use_case": "", "switched_from": "", "story": "Kimi-K3 Releases on HuggingFace 6/27"}]}], "caveats": [], "who": [], "tags": [], "use_cases": ["cheap local tool-call mapping", "llm router"], "compared": [], "makers": [{"id": 770011793, "item_id": 48125737, "url": "https://news.ycombinator.com/item?id=48125737", "by": "shreyask", "date": "2026-05-13", "version": null, "stance": "recommend", "quote": "Built a browser version of this runs entirely client-side", "excerpt": "Built a browser version of this runs entirely client-side via onnxruntime-web, no server: https://huggingface.co/spaces/shreyask/needle-playground", "context": "", "use_case": "Tool-calling model", "switched_from": "", "story": "Show HN: Needle: We Distilled Gemini Tool Calling into a 26M Model"}], "threads": [{"url": "https://news.ycombinator.com/item?id=49246804", "title": "Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=48111896", "title": "Show HN: Needle: We Distilled Gemini Tool Calling into a 26M Model", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49432317", "title": "Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49412396", "title": "I were 17, I'd learn how to build LLMs from scratch", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49322695", "title": "Models Are Getting Dumber on Purpose", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49065752", "title": "Kimi-K3 Releases on HuggingFace 6/27", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {}, "2026-03": {}, "2026-04": {}, "2026-05": {"recommend": 1}, "2026-06": {}, "2026-07": {"recommend": 1}, "2026-08": {"recommend": 4, "neutral": 1}, "2026-09": {}, "2026-10": {}}, "versions": [], "version_unnamed": 7, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 790994912, "item_id": 49437182, "url": "https://news.ycombinator.com/item?id=49437182", "by": "corysama", "date": "2026-08-25", "version": null, "stance": "neutral", "quote": "I know https://cactuscompute.com/needle is designed only to enable tool calling on tiny devices.", "excerpt": "So, I know https://cactuscompute.com/needle is designed only to enable tool calling on tiny devices. But, I wonder if anyone has used it as a CPU-side mediator between a tool and a GPU-side local LLM making semi-natural-language tool requests...", "context": "", "use_case": "", "switched_from": "", "story": "Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)"}, {"id": 790715648, "item_id": 49419728, "url": "https://news.ycombinator.com/item?id=49419728", "by": "stymaar", "date": "2026-08-24", "version": null, "stance": "recommend", "quote": "I'm a big fan of Cactus's work on the needle family of Simple Attention Network", "excerpt": "True. I'm a big fan of Cactus's work on the needle family of Simple Attention Network: https://github.com/cactus-compute/needle", "context": "", "use_case": "", "switched_from": "", "story": "I were 17, I'd learn how to build LLMs from scratch"}, {"id": 789169552, "item_id": 49323097, "url": "https://news.ycombinator.com/item?id=49323097", "by": "msdz", "date": "2026-08-16", "version": null, "stance": "recommend", "quote": "Cactus have come up with Needle [0][1], which is their tool-calling focused 14 MB model", "excerpt": "Great article. > There's a version of this future where the model card stops listing a knowledge cutoff at all, because what's left in the weights goes stale on a scale of years instead of weeks. Future? Even just recently I\u2019ve read of two approaches to this problem: Cactus have come up with Needle [0][1], which is their tool-calling focused 14 MB model (still an LLM!) \u2013 no world knowledge engrained. And instead of s", "context": "", "use_case": "", "switched_from": "", "story": "Models Are Getting Dumber on Purpose"}, {"id": 788118928, "item_id": 49257433, "url": "https://news.ycombinator.com/item?id=49257433", "by": "a022311", "date": "2026-08-11", "version": null, "stance": "recommend", "quote": "Needle and similar models solve this problem nicely", "excerpt": "Looks incredibly cool, although it ran at ~4 tok/s on my browser (I assume because I've disabled V8). I believe your numbers though. I always wanted to try using something like this to create an AI assistant for my phone. If you try to create something without AI, you'll quickly run into the problem of using NLP to map input to the equivalent of AI tool calls. On the other hand, shelling out to a full-blown LLM for e", "context": "", "use_case": "cheap local tool-call mapping", "switched_from": "", "story": "Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots"}, {"id": 788067696, "item_id": 49254231, "url": "https://news.ycombinator.com/item?id=49254231", "by": "rcarmo", "date": "2026-08-10", "version": null, "stance": "recommend", "quote": "I used Needle as a router in https://rcarmo.github.io/projects/memento/", "excerpt": "Nice. I used Needle as a router in https://rcarmo.github.io/projects/memento/, need to take a look at this", "context": "", "use_case": "LLM router", "switched_from": "", "story": "Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots"}, {"id": 785073568, "item_id": 49067098, "url": "https://news.ycombinator.com/item?id=49067098", "by": "docheinestages", "date": "2026-07-27", "version": null, "stance": "recommend", "quote": "Think of Cactus Compute's Needle", "excerpt": "Labs are focusing on creating models, small or large, that perform well on various benchmarks, including general knowledge, domain-specific expertise, and agentic capabilities. Asking for such a model while wanting to be small and fast would align with what you're describing, which I believe is different from what I'm pointing to. The model I'm describing sacrifices domain knowledge and expertise for agentic reasonin", "context": "", "use_case": "", "switched_from": "", "story": "Kimi-K3 Releases on HuggingFace 6/27"}, {"id": 769826416, "item_id": 48114151, "url": "https://news.ycombinator.com/item?id=48114151", "by": "benob", "date": "2026-05-12", "version": null, "stance": "recommend", "quote": "Deployed it to a huggingface space", "excerpt": "Deployed it to a huggingface space: https://huggingface.co/spaces/benoitfavre/needle-playground You can check the very simple docker file there.", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Needle: We Distilled Gemini Tool Calling into a 26M Model"}]}