{"slug": "bonsai", "name": "Bonsai", "category": "ai", "category_name": "AI models and assistants", "description": "", "recommenders": 3, "commenters": 6, "warners": 2, "score": 1.005, "rising": -0.98, "first": "2026-04-01", "latest": "2026-07-21", "why": "Bonsai makes the vdom incremental and it is pretty fun to work with.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 3, "neutral": 0, "compare": 1, "warn": 2}, "spark": [0, 0, 0, 0, 0, 2, 0, 0, 1, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "Bonsai makes the vdom incremental and it is pretty fun to work with.", "icon": true, "domain": null, "github_repo": "github.com/janestreet/bonsai", "official_url": "https://github.com/janestreet/bonsai", "pricing": "", "aliases": ["Bonsai", "1-Bit Bonsai", "bonsai"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 783846016, "item_id": 48990376, "url": "https://news.ycombinator.com/item?id=48990376", "by": "osener", "date": "2026-07-21", "stance": "recommend", "quote": "Bonsai makes the vdom incremental and it is pretty fun to work with.", "excerpt": "If you find this interesting, also check out their UI library called Bonsai that built on top of Incremental: https://github.com/janestreet/bonsai/ Libraries like React are pretty efficient with skipping work by using Virtual Dom, but constructing this vdom still takes time. Bonsai makes the vdom incremental and it is pretty fun to work with. I built a desktop UI library with it by targeting (now unmaintained) Revery", "context": "", "use_case": "incremental UI library", "switched_from": "", "story": "Jane Street: Incremental"}, {"id": 761579024, "item_id": 47598689, "url": "https://news.ycombinator.com/item?id=47598689", "by": "Imustaskforhelp", "date": "2026-04-01", "stance": "recommend", "quote": "its an extremely fun model to play with!", "excerpt": "I have been using @freakynit's runpod as well all be it, I like making working pomodoro apps as my own custom test, and although its not good for it (none of the prototypes work), I feel like it can be good within a specific context like Sql as you mention. I imagine this being used as sub-agents with some sota models directing them but I wasn't really able to replicate it personally (I had asked Claude to create a d", "context": "", "use_case": "small local model experiments", "switched_from": "", "story": "Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs"}, {"id": 761575728, "item_id": 47598483, "url": "https://news.ycombinator.com/item?id=47598483", "by": "nl", "date": "2026-04-01", "stance": "recommend", "quote": "I ran my custom agentic SQL debugging benchmark against it and I'm impressed.", "excerpt": "I ran my custom agentic SQL debugging benchmark against it and I'm impressed. Results: 8 passed, 0 failed, 17 errored out of 25 That puts it right between Qwen3.5-4B (7/25) and Nanbeige4.1-3B (9/25) for example, but it took only 200 seconds for the whole test. Qwen3.5 took 976 seconds and Nanbeige over 2000 (although both of these were on my 1070 so not quite the same hardware) Granite 7B 4bit does the test in 199 se", "context": "", "use_case": "agentic SQL debugging benchmark", "switched_from": "", "story": "Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 782605184, "item_id": 48912824, "url": "https://news.ycombinator.com/item?id=48912824", "by": "verdverm", "date": "2026-07-14", "stance": "warn", "quote": "Looks like they quant'd too hard at 4 bits, can't imagine the ternary being any good based on this.", "excerpt": "Preliminary analysis via lm-evaluation-harness + vllm model | disk | wikitext | gsm8k (match/error) baseline | 55G | 8.00 | 0.50/0.09 nvfp4-gptq | 27G | 8.25 | 0.47/0.9 nvfp4a16-gptq | 27G | 8.11 | 0.53/0.9 bonsai-4bit | 19G | 16.75 | 0/0 Looks like they quant'd too hard at 4 bits, can't imagine the ternary being any good based on this. Code if you'd like to reproduce or try other test sets: https://github.com/verdve", "context": "", "use_case": "", "switched_from": "", "story": "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone"}, {"id": 782601872, "item_id": 48912617, "url": "https://news.ycombinator.com/item?id=48912617", "by": "motbus3", "date": "2026-07-14", "stance": "warn", "quote": "I spent quite sometime trying to install their tools and nothing really worked.", "excerpt": "I spent quite sometime trying to install their tools and nothing really worked. I used these repos you shared but the dependencies all fail on mac", "context": "mac user", "use_case": "", "switched_from": "", "story": "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone"}]}], "who": [], "tags": [], "use_cases": ["incremental ui library", "benchmarking", "small local model experiments", "agentic sql debugging benchmark"], "compared": [{"slug": "qwen-3-5", "name": "Qwen 3.5", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [{"id": 795046464, "item_id": 49690404, "url": "https://news.ycombinator.com/item?id=49690404", "by": "jesse__", "date": "2026-09-14", "stance": "recommend", "quote": "I've been working on a voxel game engine called Bonsai for ~10 years.", "excerpt": "I've been working on a voxel game engine called Bonsai for ~10 years. Probably the most interesting thing about it at the moment is the editor. The world, and most things in it, are represented as collections of SDFs. More accurately, they're density fields, but, potato-tomato. Bonsai has undergone a large rewrite over the last couple years that's nearing completion. A world edit is defined as a bunch of SDF paramete", "context": "", "use_case": "voxel game engine", "switched_from": "", "story": "Ask HN: What are you working on? (September 2026)"}, {"id": 787870976, "item_id": 49241936, "url": "https://news.ycombinator.com/item?id=49241936", "by": "xmstan", "date": "2026-08-10", "stance": "recommend", "quote": "an open-source conversational framework for brand-safe assistants and agents", "excerpt": "I am working on Bonsai - an open-source conversational framework for brand-safe assistants and agents. It works with text and voice based channels. For example we recently built an email agent that qualifies leads and sends pdf offers back. 2 years ago (before open-sourcing) we built an AI podcaster that conducted interviews in a physical space. GH: https://github.com/utter-one/bonsai", "context": "", "use_case": "brand-safe conversational agents", "switched_from": "", "story": "Ask HN: What are you working on? (August 2026)"}, {"id": 775506496, "item_id": 48469156, "url": "https://news.ycombinator.com/item?id=48469156", "by": "nnunley", "date": "2026-06-09", "stance": "recommend", "quote": "I designed it to allow me to try to inline and reduce code for both simplifying single file examples, as well working across multiple files.", "excerpt": "I have a similar tool to shrink ray, called [bonsai](https://github.com/nnunley/bonsai). I designed it to allow me to try to inline and reduce code for both simplifying single file examples, as well working across multiple files. It uses Tree-Sitter for syntax awareness, and the [Perses algorithm](https://doi.org/10.1109/ICSE.2018.00046) as the methodology for simplification. I'd love to get some feedback if anyone's", "context": "", "use_case": "test-case reduction", "switched_from": "", "story": "Test-case reducers are underappreciated debugging tools"}], "threads": [{"url": "https://news.ycombinator.com/item?id=48910545", "title": "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=47593422", "title": "Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49686380", "title": "Ask HN: What are you working on? (September 2026)", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49233423", "title": "Ask HN: What are you working on? (August 2026)", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48987822", "title": "Jane Street: Incremental", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48459659", "title": "Test-case reducers are underappreciated debugging tools", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {}, "2026-03": {}, "2026-04": {"recommend": 2}, "2026-05": {}, "2026-06": {}, "2026-07": {"warn": 2, "compare": 1, "recommend": 1}, "2026-08": {}, "2026-09": {}, "2026-10": {}}, "quotes": [{"id": 783846016, "item_id": 48990376, "url": "https://news.ycombinator.com/item?id=48990376", "by": "osener", "date": "2026-07-21", "stance": "recommend", "quote": "Bonsai makes the vdom incremental and it is pretty fun to work with.", "excerpt": "If you find this interesting, also check out their UI library called Bonsai that built on top of Incremental: https://github.com/janestreet/bonsai/ Libraries like React are pretty efficient with skipping work by using Virtual Dom, but constructing this vdom still takes time. Bonsai makes the vdom incremental and it is pretty fun to work with. I built a desktop UI library with it by targeting (now unmaintained) Revery", "context": "", "use_case": "incremental UI library", "switched_from": "", "story": "Jane Street: Incremental"}, {"id": 782618688, "item_id": 48913668, "url": "https://news.ycombinator.com/item?id=48913668", "by": "armanj", "date": "2026-07-14", "stance": "compare", "quote": "Bonsai vs Qwen (quick) Benchmark", "excerpt": "Bonsai vs Qwen (quick) Benchmark: https://github.com/ArmanJR/PrismML-Bonsai-vs-Qwen3.5-Benchmark", "context": "", "use_case": "benchmarking", "switched_from": "", "story": "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone"}, {"id": 782605184, "item_id": 48912824, "url": "https://news.ycombinator.com/item?id=48912824", "by": "verdverm", "date": "2026-07-14", "stance": "warn", "quote": "Looks like they quant'd too hard at 4 bits, can't imagine the ternary being any good based on this.", "excerpt": "Preliminary analysis via lm-evaluation-harness + vllm model | disk | wikitext | gsm8k (match/error) baseline | 55G | 8.00 | 0.50/0.09 nvfp4-gptq | 27G | 8.25 | 0.47/0.9 nvfp4a16-gptq | 27G | 8.11 | 0.53/0.9 bonsai-4bit | 19G | 16.75 | 0/0 Looks like they quant'd too hard at 4 bits, can't imagine the ternary being any good based on this. Code if you'd like to reproduce or try other test sets: https://github.com/verdve", "context": "", "use_case": "", "switched_from": "", "story": "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone"}, {"id": 782601872, "item_id": 48912617, "url": "https://news.ycombinator.com/item?id=48912617", "by": "motbus3", "date": "2026-07-14", "stance": "warn", "quote": "I spent quite sometime trying to install their tools and nothing really worked.", "excerpt": "I spent quite sometime trying to install their tools and nothing really worked. I used these repos you shared but the dependencies all fail on mac", "context": "mac user", "use_case": "", "switched_from": "", "story": "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone"}, {"id": 761579024, "item_id": 47598689, "url": "https://news.ycombinator.com/item?id=47598689", "by": "Imustaskforhelp", "date": "2026-04-01", "stance": "recommend", "quote": "its an extremely fun model to play with!", "excerpt": "I have been using @freakynit's runpod as well all be it, I like making working pomodoro apps as my own custom test, and although its not good for it (none of the prototypes work), I feel like it can be good within a specific context like Sql as you mention. I imagine this being used as sub-agents with some sota models directing them but I wasn't really able to replicate it personally (I had asked Claude to create a d", "context": "", "use_case": "small local model experiments", "switched_from": "", "story": "Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs"}, {"id": 761575728, "item_id": 47598483, "url": "https://news.ycombinator.com/item?id=47598483", "by": "nl", "date": "2026-04-01", "stance": "recommend", "quote": "I ran my custom agentic SQL debugging benchmark against it and I'm impressed.", "excerpt": "I ran my custom agentic SQL debugging benchmark against it and I'm impressed. Results: 8 passed, 0 failed, 17 errored out of 25 That puts it right between Qwen3.5-4B (7/25) and Nanbeige4.1-3B (9/25) for example, but it took only 200 seconds for the whole test. Qwen3.5 took 976 seconds and Nanbeige over 2000 (although both of these were on my 1070 so not quite the same hardware) Granite 7B 4bit does the test in 199 se", "context": "", "use_case": "agentic SQL debugging benchmark", "switched_from": "", "story": "Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs"}]}