{"slug": "qwen-3-6", "name": "Qwen 3.6", "category": "ai", "category_name": "AI models and assistants", "description": "Alibaba's open-weight LLM family, including 27B dense and 35B-A3B mixture-of-experts models.", "recommenders": 93, "commenters": 142, "warners": 10, "score": 83.97, "rising": -1.89, "first": "2026-04-16", "latest": "2026-10-20", "why": "Liked as a capable model that runs locally on consumer GPUs and Macs and is many commenters' daily driver for coding and everyday tasks.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 97, "neutral": 39, "compare": 5, "warn": 10}, "spark": [0, 0, 0, 0, 0, 6, 6, 51, 12, 15, 6, 1], "listed": true, "loved": ["Daily driver", "Consumer hardware"], "caveat": "Below frontier", "short_quote": "I genuinely think this is the first consumer hardware-grade model that can actually replace frontiers for a lot of workloads.", "icon": true, "domain": "qwen.ai", "github_repo": null, "official_url": "https://qwen.ai", "pricing": "open source", "aliases": ["Qwen3.6", "Qwen 3.6 27B", "Qwen 3.6", "Qwen3.6-27B", "Qwen3.6 27B", "Qwen3.6-35B-A3B", "Qwen 3.6 35B A3B", "Qwen 3.6 35B-A3B", "Qwen3.6 35B A3B", "Qwen3.6 35B", "Qwen 3.6 27b", "Qwen 3.6 35B", "Qwen3.6-27b", "Qwen3.6-35B", "Qwen3.6-35B-A3B-MTP", "Qwen3.6-35b", "Qwen3.6 35B-A3B", "Qwen3.6 35b", "Qwen 3.6-35B-A3B", "Qwen-3.6", "Qwen3.6 27b", "Qwen3.6-35b-a3b", "Qwen3.6-35B A3B"], "verdict": [{"text": "Liked as a capable model that runs locally on consumer GPUs and Macs", "cites": [{"id": 791579152, "item_id": 49473697, "url": "https://news.ycombinator.com/item?id=49473697", "by": "trucks-refinish", "date": "2026-08-28", "stance": "recommend", "quote": "It's a very capable model, not as smart as a frontier model but it is still very useful.", "excerpt": "I'm able to run qwen 3.6 35b a3b with my rtx 2060 super which also has 8gb of vram. It's a very capable model, not as smart as a frontier model but it is still very useful. I use an 6bit quant and get around 20 tokens per second.", "context": "runs on an RTX 2060 Super with 8GB VRAM", "use_case": "local LLM on 8GB GPU", "switched_from": "", "story": "Small Models Have Arrived"}, {"id": 792403488, "item_id": 49525218, "url": "https://news.ycombinator.com/item?id=49525218", "by": "prometheus1992", "date": "2026-09-01", "stance": "recommend", "quote": "I am running Qwen3.6-35B-A3B on my 16GB M3 and get 7-8 tokens/sec", "excerpt": "It's hard to believe 16GB unified memory will give you 5 tok/sec unless you are ignoring the thermal warnings. I am running Qwen3.6-35B-A3B on my 16GB M3 and get 7-8 tokens/sec with all the optimizations while keeping the peak memory and thermal warnings at check. https://github.com/deepanwadhwa/samosa-chat", "context": "", "use_case": "local LLM inference", "switched_from": "", "story": "Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/s"}, {"id": 783464064, "item_id": 48966504, "url": "https://news.ycombinator.com/item?id=48966504", "by": "nsbk", "date": "2026-07-19", "stance": "recommend", "quote": "a single 3090 already provides a great experience at a reasonable quant and context size", "excerpt": "I run a 2x 3090 rig, but a single 3090 already provides a great experience at a reasonable quant and context size. On a single card I used to run Qwen_Qwen3.6-27B-Q4_K_M or similarly quantized 35B MoE at 65536 context size", "context": "runs 2x 3090 rig", "use_case": "local inference on 3090", "switched_from": "", "story": "Qwen3.8 is launching and going open-weight soon"}, {"id": 779606528, "item_id": 48725408, "url": "https://news.ycombinator.com/item?id=48725408", "by": "cpburns2009", "date": "2026-06-29", "stance": "recommend", "quote": "A 32gb card does run it nicely.", "excerpt": "A 32gb card does run it nicely. I use unsloth's UD-Q5_K_XL at 256k context (k/v at q8_0), and get ~67 t/s on a 5090. I still need to look into MTP.", "context": "", "use_case": "local LLM on a 5090", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}]}, {"text": " and is many commenters' daily driver for coding and everyday tasks.", "cites": [{"id": 786414864, "item_id": 49150929, "url": "https://news.ycombinator.com/item?id=49150929", "by": "nozzlegear", "date": "2026-08-02", "stance": "recommend", "quote": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April.", "excerpt": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April. The Qwen3.6 line is easily the best local model I've tried, and I've tried a lot. I've got it diligently grinding away on my laptop right now, reviewing and fixing some bugs in my F# code.", "context": "", "use_case": "local coding and bug fixing", "switched_from": "Claude", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 789246176, "item_id": 49327886, "url": "https://news.ycombinator.com/item?id=49327886", "by": "digikata", "date": "2026-08-17", "stance": "recommend", "quote": "3.6 27B was easily a daily driver with only an occasional need to pop up to larger models for planning.", "excerpt": "Qwen3.6 27B is very usable, and dialing back thinking modes woth Qwen3.8 bring it close, but 3.8 stills feels slower. Unknown to me if the results are qualitatively better or worse overall - with the heavier thinking 3.8 felt worse in terms of coding tasks, but I think Im comparing a newly released model to one that has had a lot of harness tuning. 3.6 27B was easily a daily driver with only an occasional need to pop", "context": "", "use_case": "daily coding driver", "switched_from": "", "story": "Qwen 3.8 27B is excellent, but it defaults to overthinking things"}, {"id": 778616800, "item_id": 48663550, "url": "https://news.ycombinator.com/item?id=48663550", "by": "crazycracker", "date": "2026-06-24", "stance": "recommend", "quote": "I use qwen3.6:27b as a daily driver and I am heavily impressed with it.", "excerpt": "Historically the cost of compute has also gone down. Like just look at it as compared to a year ago. We have amazing open source models that can run on consumer hardware and if we go away from our obsession of using opus 4.8 or mythos for everything then it actually is super amazing to see what these open source models could do. I use qwen3.6:27b as a daily driver and I am heavily impressed with it.", "context": "", "use_case": "daily local model", "switched_from": "", "story": "For Most of the World, Open-Source AI Is the Only Way Forward"}, {"id": 797379840, "item_id": 49836240, "url": "https://news.ycombinator.com/item?id=49836240", "by": "LeBit", "date": "2026-09-24", "stance": "recommend", "quote": "I find them quite useful!", "excerpt": "I use the following models: - Qwen3.5 9B - Qwen3.6 35B A3B - Qwen3.8 27B - Gemma 4 26B A4B - Gemma 31B - Muse Glimmer 30B I have 48G. The MacMini is used solely for inference. llama.cpp + llama-swap. People are saying local models are crap and serves no purposes. I use them to help me spell check, write emails, write text messages, write JIRA tickets, write PR comments, etc. I also use them in coding agents to comple", "context": "Mac Mini with 48G", "use_case": "local inference for writing and coding", "switched_from": "", "story": "Best LLM for every budget, updated daily"}, {"id": 779581937, "item_id": 48723871, "url": "https://news.ycombinator.com/item?id=48723871", "by": "Zambyte", "date": "2026-06-29", "stance": "recommend", "quote": "I have been very blown away by how well it works.", "excerpt": "I have been using pi (and previously the codex cli) with Qwen 3.6 27b with 100k context for my development at work, and I have been very blown away by how well it works. It's not perfect, but it's enough to accelerate my normal development flow. I mostly use it for writing Go and C#.", "context": "", "use_case": "local coding with Go and C#", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}]}, {"text": " The most common caveat is that it falls short of frontier models on coherence and harder coding work", "cites": [{"id": 799810306, "item_id": 49988144, "url": "https://news.ycombinator.com/item?id=49988144", "by": "colingauvin", "date": "2026-10-07", "stance": "warn", "quote": "the coherence was really not great compared to frontier", "excerpt": "The author's configuration is..... Interesting. I don't disagree with the title. This spring folks were saying \"I use Qwen 3.6 27B (or 35 A3B) as a daily driver\". They either had low standards or were lying. The models were good enough technically, they capable enough, that you could see the potential, but the coherence was really not great compared to frontier. They'd lose track of what they were doing, mix things u", "context": "runs models on 2x Sparks", "use_case": "local daily driver model", "switched_from": "", "story": "Local AI is real now. And it's blowing my mind"}, {"id": 779670192, "item_id": 48729387, "url": "https://news.ycombinator.com/item?id=48729387", "by": "PeterStuer", "date": "2026-06-29", "stance": "warn", "quote": "for software development it is unsurprisingly a dramatic downgrade from the Opus I use as my daily driver.", "excerpt": "Been running it on a 9950x3D with 96GB and a 4090. Speedwise it is fine. Bit while not completely useless, for software development it is unsurprisingly a dramatic downgrade from the Opus I use as my daily driver.", "context": "", "use_case": "local software development", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 791579152, "item_id": 49473697, "url": "https://news.ycombinator.com/item?id=49473697", "by": "trucks-refinish", "date": "2026-08-28", "stance": "recommend", "quote": "It's a very capable model, not as smart as a frontier model but it is still very useful.", "excerpt": "I'm able to run qwen 3.6 35b a3b with my rtx 2060 super which also has 8gb of vram. It's a very capable model, not as smart as a frontier model but it is still very useful. I use an 6bit quant and get around 20 tokens per second.", "context": "runs on an RTX 2060 Super with 8GB VRAM", "use_case": "local LLM on 8GB GPU", "switched_from": "", "story": "Small Models Have Arrived"}, {"id": 779649472, "item_id": 48728092, "url": "https://news.ycombinator.com/item?id=48728092", "by": "m3kw9", "date": "2026-06-29", "stance": "warn", "quote": "it can't get past a simple coding test that 5.5 passes with light reasoning", "excerpt": "Hmm, i used it and it can't get past a simple coding test that 5.5 passes with light reasoning", "context": "", "use_case": "coding", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 785947718, "item_id": 49121732, "url": "https://news.ycombinator.com/item?id=49121732", "by": "user43928", "date": "2026-07-31", "stance": "warn", "quote": "\"I love hallucinated garbage\": Sonnet, Qwen 3.6, GPT 5.4 mini, GPT 5.3 Codex", "excerpt": "In my mind there are three tiers: The SOTA: Fable, GPT 5.6 Sol, Opus 5 The \"enterprise admin did not turn on the new models\": Opus 4.8, GPT 5.5 The \"I love hallucinated garbage\": Sonnet, Qwen 3.6, GPT 5.4 mini, GPT 5.3 Codex, etc. Results vary widely", "context": "", "use_case": "", "switched_from": "", "story": "Google fixed more Chrome bugs in June than over the past two years, thanks to AI"}]}, {"text": ", and some report runaway thinking loops and failed tool calls.", "cites": [{"id": 791374224, "item_id": 49460889, "url": "https://news.ycombinator.com/item?id=49460889", "by": "Balinares", "date": "2026-08-27", "stance": "warn", "quote": "but those do exit eventually, unlike with Qwen 3.6.", "excerpt": "I've found it tends toward long thinking loops even for simple tasks (and any quantization seems to increase their length), but those do exit eventually, unlike with Qwen 3.6. I use the Unsloth UD_Q2_K_XL GGUF with default parameters, along with that custom template linked elsewhere in the thread, and no K/V cache quantization.", "context": "", "use_case": "local LLM", "switched_from": "", "story": "GLM-5.3-Flash"}, {"id": 787992352, "item_id": 49249522, "url": "https://news.ycombinator.com/item?id=49249522", "by": "bitexploder", "date": "2026-08-10", "stance": "recommend", "quote": "My final 35B A3B tip -- use an extremely simple harness.", "excerpt": "I have a custom A3B proxy that caps its thinking off. It is a known issue with the model that Qwen themselves documented but is almost never addressed in any harnesses. I also patched up a few other known bugs in the proxy. I highly recommend you shim A3B and when it hits 2K thinking tokens inject (paraphrasing) 'Time to wrap it up bud! Get to work' into its thinking stream. It almost always gets to work. If it needs", "context": "runs a custom proxy for local models", "use_case": "local agentic coding", "switched_from": "", "story": "Meta Muse Glimmer \u2013 open weights 30B local coding model"}, {"id": 782152256, "item_id": 48884516, "url": "https://news.ycombinator.com/item?id=48884516", "by": "docheinestages", "date": "2026-07-12", "stance": "warn", "quote": "I tried Qwen 3.6 27B, but I can only get around 10 tokens per second, but it's painfully slow, and it often fails during `write_file` tool calls", "excerpt": "I'm on a 48 GB M5 Macbook Pro. I use 4-bit quants with a context window of 16-32k. I tried Qwen 3.6 27B, but I can only get around 10 tokens per second, but it's painfully slow, and it often fails during `write_file` tool calls, even with Qwen Code.", "context": "48 GB M5 Macbook Pro user", "use_case": "local coding model", "switched_from": "", "story": "Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k"}, {"id": 779714784, "item_id": 48732174, "url": "https://news.ycombinator.com/item?id=48732174", "by": "ThunderSizzle", "date": "2026-06-30", "stance": "recommend", "quote": "35B is pretty good", "excerpt": "I'll give 27B-MTP a try. I think I can tolerate 45 tps if the results are technically better. 35B is pretty good, but definitely shows it's inabilities at times (probably either due to the heavy caching quantization I'm doing, or the heavy model quantization vs what 2 GPUs could run). My biggest gripe is that both pi and opencode seem to have trouble parsing the thinking blocks at times, and the model sometimes cuts-", "context": "runs local models on 2 GPUs", "use_case": "local coding model", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}]}], "praise": [{"label": "Daily driver", "summary": "Many use it as their main model for coding and general work.", "quotes": [{"id": 786414864, "item_id": 49150929, "url": "https://news.ycombinator.com/item?id=49150929", "by": "nozzlegear", "date": "2026-08-02", "stance": "recommend", "quote": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April.", "excerpt": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April. The Qwen3.6 line is easily the best local model I've tried, and I've tried a lot. I've got it diligently grinding away on my laptop right now, reviewing and fixing some bugs in my F# code.", "context": "", "use_case": "local coding and bug fixing", "switched_from": "Claude", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 789246176, "item_id": 49327886, "url": "https://news.ycombinator.com/item?id=49327886", "by": "digikata", "date": "2026-08-17", "stance": "recommend", "quote": "3.6 27B was easily a daily driver with only an occasional need to pop up to larger models for planning.", "excerpt": "Qwen3.6 27B is very usable, and dialing back thinking modes woth Qwen3.8 bring it close, but 3.8 stills feels slower. Unknown to me if the results are qualitatively better or worse overall - with the heavier thinking 3.8 felt worse in terms of coding tasks, but I think Im comparing a newly released model to one that has had a lot of harness tuning. 3.6 27B was easily a daily driver with only an occasional need to pop", "context": "", "use_case": "daily coding driver", "switched_from": "", "story": "Qwen 3.8 27B is excellent, but it defaults to overthinking things"}, {"id": 779581937, "item_id": 48723871, "url": "https://news.ycombinator.com/item?id=48723871", "by": "Zambyte", "date": "2026-06-29", "stance": "recommend", "quote": "I have been very blown away by how well it works.", "excerpt": "I have been using pi (and previously the codex cli) with Qwen 3.6 27b with 100k context for my development at work, and I have been very blown away by how well it works. It's not perfect, but it's enough to accelerate my normal development flow. I mostly use it for writing Go and C#.", "context": "", "use_case": "local coding with Go and C#", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}]}, {"label": "Consumer hardware", "summary": "Runs usefully on modest GPUs and Macs, including 8GB VRAM cards and 16GB laptops.", "quotes": [{"id": 791579152, "item_id": 49473697, "url": "https://news.ycombinator.com/item?id=49473697", "by": "trucks-refinish", "date": "2026-08-28", "stance": "recommend", "quote": "It's a very capable model, not as smart as a frontier model but it is still very useful.", "excerpt": "I'm able to run qwen 3.6 35b a3b with my rtx 2060 super which also has 8gb of vram. It's a very capable model, not as smart as a frontier model but it is still very useful. I use an 6bit quant and get around 20 tokens per second.", "context": "runs on an RTX 2060 Super with 8GB VRAM", "use_case": "local LLM on 8GB GPU", "switched_from": "", "story": "Small Models Have Arrived"}, {"id": 792403488, "item_id": 49525218, "url": "https://news.ycombinator.com/item?id=49525218", "by": "prometheus1992", "date": "2026-09-01", "stance": "recommend", "quote": "I am running Qwen3.6-35B-A3B on my 16GB M3 and get 7-8 tokens/sec", "excerpt": "It's hard to believe 16GB unified memory will give you 5 tok/sec unless you are ignoring the thermal warnings. I am running Qwen3.6-35B-A3B on my 16GB M3 and get 7-8 tokens/sec with all the optimizations while keeping the peak memory and thermal warnings at check. https://github.com/deepanwadhwa/samosa-chat", "context": "", "use_case": "local LLM inference", "switched_from": "", "story": "Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/s"}, {"id": 783464064, "item_id": 48966504, "url": "https://news.ycombinator.com/item?id=48966504", "by": "nsbk", "date": "2026-07-19", "stance": "recommend", "quote": "a single 3090 already provides a great experience at a reasonable quant and context size", "excerpt": "I run a 2x 3090 rig, but a single 3090 already provides a great experience at a reasonable quant and context size. On a single card I used to run Qwen_Qwen3.6-27B-Q4_K_M or similarly quantized 35B MoE at 65536 context size", "context": "runs 2x 3090 rig", "use_case": "local inference on 3090", "switched_from": "", "story": "Qwen3.8 is launching and going open-weight soon"}]}, {"label": "Speed", "summary": "The 35B-A3B MoE variant in particular reaches high token rates locally.", "quotes": [{"id": 787905568, "item_id": 49244098, "url": "https://news.ycombinator.com/item?id=49244098", "by": "Roark66", "date": "2026-08-10", "stance": "recommend", "quote": "The moe works well in Librechat combined with my local search/Web retrieval system.", "excerpt": "I've been using Qwen3.6 models locally for a couple of weeks. Both the A3B moe and the dense variant. The moe works well in Librechat combined with my local search/Web retrieval system. All components use open source projects such as SearXNG, Crawl4AI, MetaMCP, Jina rerank, but all needed quite a bit of coding to work nicely together. I get 140 tok/s on short prompts on an rtx3090 on the qwen3.6 moe which makes is ea", "context": "", "use_case": "local web research and planning", "switched_from": "", "story": "DeepSeek costs OpenCode Go user $1.14/day; dual DGX breaks even in 24 years"}, {"id": 781810977, "item_id": 48863186, "url": "https://news.ycombinator.com/item?id=48863186", "by": "syntaxing", "date": "2026-07-10", "stance": "recommend", "quote": "Been using Qwen3.6-35B @ Q_8 as my main driver and it\u2019s been great with 60 TPS for generation.", "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": "main local model", "switched_from": "", "story": "Hands-On with the AMD Ryzen AI Halo"}, {"id": 797400369, "item_id": 49837523, "url": "https://news.ycombinator.com/item?id=49837523", "by": "sznio", "date": "2026-09-24", "stance": "recommend", "quote": "it's both much faster and much more capable than Haiku", "excerpt": "Nothing, because current Haiku is useless. But I use Qwen3.6 35B-A3B and it's both much faster and much more capable than Haiku. And Qwen3.8 27B can do so much more, decently replaces Sonnet, albeit not as fast. Luna doesn't feel more capable nor faster than Qwen3.8. I think Anthropic could surprise us with some more innovation in this niche.", "context": "", "use_case": "", "switched_from": "Claude Haiku", "story": "Claude Opus 5.5"}]}, {"label": "Replaces paid APIs", "summary": "Some commenters dropped Claude subscriptions or use it to save cloud tokens.", "quotes": [{"id": 786414864, "item_id": 49150929, "url": "https://news.ycombinator.com/item?id=49150929", "by": "nozzlegear", "date": "2026-08-02", "stance": "recommend", "quote": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April.", "excerpt": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April. The Qwen3.6 line is easily the best local model I've tried, and I've tried a lot. I've got it diligently grinding away on my laptop right now, reviewing and fixing some bugs in my F# code.", "context": "", "use_case": "local coding and bug fixing", "switched_from": "Claude", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786523904, "item_id": 49157744, "url": "https://news.ycombinator.com/item?id=49157744", "by": "jkingsman", "date": "2026-08-03", "stance": "recommend", "quote": "I use it for personal tasks that are just not worth the claude tokens", "excerpt": "Not OP, but I use it for personal tasks that are just not worth the claude tokens -- rooting through historical medical records to unify prescription history, super-OCR'ing thousands of PDF pages (i.e. beyond PDF dumping -- vision means it can look at tables, understand tricky things like a continuation of a block quote or aside on the next page, etc.), and when power is cheap I'll just let it noodle on little projec", "context": "runs models locally", "use_case": "local OCR and personal data tasks", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 780244992, "item_id": 48765312, "url": "https://news.ycombinator.com/item?id=48765312", "by": "pyreko", "date": "2026-07-02", "stance": "recommend", "quote": "been happy throwing Qwen3.6 on my old MBP", "excerpt": "Same here, been happy throwing Qwen3.6 on my old MBP - no it's not as fast as Claude which I use at work, but it works well enough locally and I don't have to worry about credits or shit like the rug getting pulled under me in terms of capabilities.", "context": "", "use_case": "local LLM", "switched_from": "", "story": "Kimi K2.7 Code is generally available in GitHub Copilot"}]}, {"label": "Agents and tools", "summary": "Works well in agent setups, eagerly using web search and orchestrating other agents.", "quotes": [{"id": 778007008, "item_id": 48625438, "url": "https://news.ycombinator.com/item?id=48625438", "by": "sleepyeldrazi", "date": "2026-06-21", "stance": "recommend", "quote": "That's why I like qwen3.6 27B, it has 0 ego, it knows that it doesn't have complete world knowledge", "excerpt": "That's why I like qwen3.6 27B, it has 0 ego, it knows that it doesn't have complete world knowledge, so when it sees a web_search tool it searches all the time. Even qwen3.5 9B is mostly search-eager (but given the size, it's weaker on reasoning on the results if that's needed). I use a stock pi harness with only web_search and web_fetch (cleans up the html to only keep text) tools defined. I have given up on making ", "context": "", "use_case": "web search and retrieval", "switched_from": "", "story": null}, {"id": 787438721, "item_id": 49214920, "url": "https://news.ycombinator.com/item?id=49214920", "by": "nozzlegear", "date": "2026-08-07", "stance": "recommend", "quote": "Then I have Qwen3.6-35B-A3B as the \"orchestrator\" agent (it's just what I use for my daily driver)", "excerpt": "Yeah, they're excellent. I use a small Qwen3-4B-Instruct as the \"actor\" agent which goes out to the website, searches for a batch of items, decides which best matches the criteria, and puts them in the cart. Then I have Qwen3.6-35B-A3B as the \"orchestrator\" agent (it's just what I use for my daily driver), it's responsible for chatting with my wife about the menu, recipes and ingredients, and tells the smaller agent ", "context": "", "use_case": "orchestrating shopping agent", "switched_from": "", "story": "Kitesurf: Agent-first browser that runs in V8 isolates"}, {"id": 778303008, "item_id": 48643938, "url": "https://news.ycombinator.com/item?id=48643938", "by": "bakies", "date": "2026-06-23", "stance": "recommend", "quote": "It's the first local model that's actually working for me in a coding agent I've tried.", "excerpt": "I've just started using qwen3.6:35b a couple days ago running on my framework desktop and rather impressed. It runs really well and reminds me of probably the first Claude model I used. It's the first local model that's actually working for me in a coding agent I've tried. Very exciting!", "context": "", "use_case": "local coding agent model", "switched_from": "", "story": "VibeThinker: 3B param model that beats Opus 4.5 on reasoning with novel SFT+GRPO"}]}], "caveats": [{"label": "Below frontier", "summary": "Commenters find it less coherent and capable than frontier models for serious development.", "quotes": [{"id": 799810306, "item_id": 49988144, "url": "https://news.ycombinator.com/item?id=49988144", "by": "colingauvin", "date": "2026-10-07", "stance": "warn", "quote": "the coherence was really not great compared to frontier", "excerpt": "The author's configuration is..... Interesting. I don't disagree with the title. This spring folks were saying \"I use Qwen 3.6 27B (or 35 A3B) as a daily driver\". They either had low standards or were lying. The models were good enough technically, they capable enough, that you could see the potential, but the coherence was really not great compared to frontier. They'd lose track of what they were doing, mix things u", "context": "runs models on 2x Sparks", "use_case": "local daily driver model", "switched_from": "", "story": "Local AI is real now. And it's blowing my mind"}, {"id": 779670192, "item_id": 48729387, "url": "https://news.ycombinator.com/item?id=48729387", "by": "PeterStuer", "date": "2026-06-29", "stance": "warn", "quote": "for software development it is unsurprisingly a dramatic downgrade from the Opus I use as my daily driver.", "excerpt": "Been running it on a 9950x3D with 96GB and a 4090. Speedwise it is fine. Bit while not completely useless, for software development it is unsurprisingly a dramatic downgrade from the Opus I use as my daily driver.", "context": "", "use_case": "local software development", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 779649472, "item_id": 48728092, "url": "https://news.ycombinator.com/item?id=48728092", "by": "m3kw9", "date": "2026-06-29", "stance": "warn", "quote": "it can't get past a simple coding test that 5.5 passes with light reasoning", "excerpt": "Hmm, i used it and it can't get past a simple coding test that 5.5 passes with light reasoning", "context": "", "use_case": "coding", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}]}, {"label": "Thinking loops", "summary": "It can spin in long thinking loops that some cap with a proxy.", "quotes": [{"id": 791374224, "item_id": 49460889, "url": "https://news.ycombinator.com/item?id=49460889", "by": "Balinares", "date": "2026-08-27", "stance": "warn", "quote": "but those do exit eventually, unlike with Qwen 3.6.", "excerpt": "I've found it tends toward long thinking loops even for simple tasks (and any quantization seems to increase their length), but those do exit eventually, unlike with Qwen 3.6. I use the Unsloth UD_Q2_K_XL GGUF with default parameters, along with that custom template linked elsewhere in the thread, and no K/V cache quantization.", "context": "", "use_case": "local LLM", "switched_from": "", "story": "GLM-5.3-Flash"}, {"id": 787992352, "item_id": 49249522, "url": "https://news.ycombinator.com/item?id=49249522", "by": "bitexploder", "date": "2026-08-10", "stance": "recommend", "quote": "My final 35B A3B tip -- use an extremely simple harness.", "excerpt": "I have a custom A3B proxy that caps its thinking off. It is a known issue with the model that Qwen themselves documented but is almost never addressed in any harnesses. I also patched up a few other known bugs in the proxy. I highly recommend you shim A3B and when it hits 2K thinking tokens inject (paraphrasing) 'Time to wrap it up bud! Get to work' into its thinking stream. It almost always gets to work. If it needs", "context": "runs a custom proxy for local models", "use_case": "local agentic coding", "switched_from": "", "story": "Meta Muse Glimmer \u2013 open weights 30B local coding model"}]}, {"label": "Tool-call glitches", "summary": "Some see failed tool calls, mid-thinking cutoffs or parsing problems, partly fixed by improved templates.", "quotes": [{"id": 782152256, "item_id": 48884516, "url": "https://news.ycombinator.com/item?id=48884516", "by": "docheinestages", "date": "2026-07-12", "stance": "warn", "quote": "I tried Qwen 3.6 27B, but I can only get around 10 tokens per second, but it's painfully slow, and it often fails during `write_file` tool calls", "excerpt": "I'm on a 48 GB M5 Macbook Pro. I use 4-bit quants with a context window of 16-32k. I tried Qwen 3.6 27B, but I can only get around 10 tokens per second, but it's painfully slow, and it often fails during `write_file` tool calls, even with Qwen Code.", "context": "48 GB M5 Macbook Pro user", "use_case": "local coding model", "switched_from": "", "story": "Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k"}, {"id": 779714784, "item_id": 48732174, "url": "https://news.ycombinator.com/item?id=48732174", "by": "ThunderSizzle", "date": "2026-06-30", "stance": "recommend", "quote": "35B is pretty good", "excerpt": "I'll give 27B-MTP a try. I think I can tolerate 45 tps if the results are technically better. 35B is pretty good, but definitely shows it's inabilities at times (probably either due to the heavy caching quantization I'm doing, or the heavy model quantization vs what 2 GPUs could run). My biggest gripe is that both pi and opencode seem to have trouble parsing the thinking blocks at times, and the model sometimes cuts-", "context": "runs local models on 2 GPUs", "use_case": "local coding model", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 789218880, "item_id": 49326180, "url": "https://news.ycombinator.com/item?id=49326180", "by": "hedgehog", "date": "2026-08-16", "stance": "recommend", "quote": "I use Qwen 3.6 a fair amount using his template and it fixes some issues I saw with the upstream versions.", "excerpt": "Look for improved templates by \"froggeric\" on Hugging Face. I use Qwen 3.6 a fair amount using his template and it fixes some issues I saw with the upstream versions.", "context": "", "use_case": "local model", "switched_from": "", "story": "Qwen 3.8 27B is excellent, but it defaults to overthinking things"}]}, {"label": "Needs precise specs", "summary": "It struggles with vague specs and needs detailed upfront guidance or split sessions.", "quotes": [{"id": 778265152, "item_id": 48641572, "url": "https://news.ycombinator.com/item?id=48641572", "by": "ekidd", "date": "2026-06-23", "stance": "recommend", "quote": "This is a fun little model that's smart enough to do debugging, refactoring, and implementing \"clean\" specs", "excerpt": "A GPU with 24GBs of RAM is mostly useful for running a very carefully squeezed Qwen3.6 27B (4-bit Unsloth quants, 8-bit K/V cache, possibly MTP, 128k context). This is a fun little model that's smart enough to do debugging, refactoring, and implementing \"clean\" specs that don't force it to make complicated design choices. I've seen it rip through a 9-year-old Terraform AWS config, and (without using the network) corr", "context": "runs local models on a 24GB GPU", "use_case": "local debugging and refactoring", "switched_from": "", "story": "Unsloth GLM-5.2 \u2013 How to Run Locally"}, {"id": 786457408, "item_id": 49153588, "url": "https://news.ycombinator.com/item?id=49153588", "by": "cybertim", "date": "2026-08-03", "stance": "recommend", "quote": "I also \"evolved\" into 27b (q8 unsloth)", "excerpt": "I also \"evolved\" into 27b (q8 unsloth) and pi.dev (tried many combinations) feels for me the same as opus 4.5 that i use at work, faster even (using 2x 3080 20GB gives me 60-80tk/s). Though you do need to feed it more details up front (about what exactly you are planning to do and a good written skill.md) but I work that way anyways, im hyped for 3.8", "context": "", "use_case": "local coding model", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 780548112, "item_id": 48784257, "url": "https://news.ycombinator.com/item?id=48784257", "by": "aka-rider", "date": "2026-07-04", "stance": "recommend", "quote": "I run Qwen3.6 on RTX4090, and it does amazing job for the most parts.", "excerpt": "I run Qwen3.6 on RTX4090, and it does amazing job for the most parts. For coding task, one needs to break the session among multiple calls I made https://github.com/aka-rider/orqestra but it's possible to do the same in almost any modern harness directly. The main idea is: - separate session that burns context on reading code and calling tools (context7, etc) -> markdown report \"here are relevant patrts of code, docs", "context": "local LLM user", "use_case": "local coding", "switched_from": "", "story": "Jamesob's guide to running SOTA LLMs locally"}]}, {"label": "Laptop heat", "summary": "Running it on a laptop during heavy work causes heat and fan noise.", "quotes": [{"id": 779582064, "item_id": 48723879, "url": "https://news.ycombinator.com/item?id=48723879", "by": "iagooar", "date": "2026-06-29", "stance": "recommend", "quote": "I love my MacBook Pro M5 128GB RAM and I love qwen3.6.", "excerpt": "I love my MacBook Pro M5 128GB RAM and I love qwen3.6. BUT DO NOT buy this MacBook if you plan on doing serious coding using local LLMs with it. The reason is simple: your fingers will burn and your head will explode from the noise. Running any kind of sophisticated job on the very laptop you are using is just not viable. Sure you can use it in clamshell mode, but forget touching it while working with AI coding or ag", "context": "", "use_case": "local coding", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}]}], "who": [{"by": "apatheticonion", "context": "maintains OSS GitHub repos", "url": "https://news.ycombinator.com/item?id=50173430", "date": "2026-10-20"}, {"by": "LeBit", "context": "Mac Mini with 48G", "url": "https://news.ycombinator.com/item?id=49836240", "date": "2026-09-24"}, {"by": "abalashov", "context": "individual contributor", "url": "https://news.ycombinator.com/item?id=49815221", "date": "2026-09-23"}, {"by": "trucks-refinish", "context": "runs on an RTX 2060 Super with 8GB VRAM", "url": "https://news.ycombinator.com/item?id=49473697", "date": "2026-08-28"}, {"by": "bitexploder", "context": "runs a custom proxy for local models", "url": "https://news.ycombinator.com/item?id=49249522", "date": "2026-08-10"}, {"by": "h4fizwasabie", "context": "runs 4GB VRAM", "url": "https://news.ycombinator.com/item?id=49227405", "date": "2026-08-08"}, {"by": "jkingsman", "context": "runs models locally", "url": "https://news.ycombinator.com/item?id=49157744", "date": "2026-08-03"}, {"by": "AbsurdCensor", "context": "runs local AI box on solar", "url": "https://news.ycombinator.com/item?id=49156910", "date": "2026-08-03"}, {"by": "badsectoracula", "context": "runs local LLMs on a 24GB VRAM GPU", "url": "https://news.ycombinator.com/item?id=49115614", "date": "2026-07-30"}, {"by": "nyrikki", "context": "30 years of experience", "url": "https://news.ycombinator.com/item?id=49049803", "date": "2026-07-25"}, {"by": "nsbk", "context": "runs 2x 3090 rig", "url": "https://news.ycombinator.com/item?id=48966504", "date": "2026-07-19"}, {"by": "lebovic", "context": "RL and finetuning practitioner", "url": "https://news.ycombinator.com/item?id=48929886", "date": "2026-07-15"}], "tags": ["self-hostable", "for coding agents", "for consumer GPUs", "for Apple Silicon Macs", "for fine-tuning"], "use_cases": ["local coding", "local coding model", "local llm", "local model", "local inference", "local llm inference"], "compared": [{"slug": "llama-cpp", "name": "llama.cpp", "co": 35, "switched_from": 0, "switched_to": 0}, {"slug": "pi", "name": "Pi", "co": 24, "switched_from": 0, "switched_to": 0}, {"slug": "opencode", "name": "OpenCode", "co": 20, "switched_from": 0, "switched_to": 0}, {"slug": "gemma", "name": "Gemma", "co": 19, "switched_from": 0, "switched_to": 0}, {"slug": "qwen-3-5", "name": "Qwen 3.5", "co": 15, "switched_from": 2, "switched_to": 0}, {"slug": "claude-opus", "name": "Claude Opus", "co": 18, "switched_from": 0, "switched_to": 0}, {"slug": "claude", "name": "Claude", "co": 16, "switched_from": 1, "switched_to": 0}, {"slug": "qwen-3-8", "name": "Qwen 3.8", "co": 16, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=48542100", "title": "Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?", "mentions": 26}, {"url": "https://news.ycombinator.com/item?id=48555993", "title": "Running local models is good now", "mentions": 13}, {"url": "https://news.ycombinator.com/item?id=48721903", "title": "Qwen 3.6 27B is the sweet spot for local development", "mentions": 12}, {"url": "https://news.ycombinator.com/item?id=49150470", "title": "Qwen3.8-Max: A New Bar for Coding and Cowork", "mentions": 6}, {"url": "https://news.ycombinator.com/item?id=49324985", "title": "Qwen 3.8 27B is excellent, but it defaults to overthinking things", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=48775921", "title": "Jamesob's guide to running SOTA LLMs locally", "mentions": 3}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {}, "2026-03": {}, "2026-04": {"recommend": 6, "warn": 1, "neutral": 1}, "2026-05": {"recommend": 6, "neutral": 2}, "2026-06": {"recommend": 51, "warn": 5, "compare": 4, "neutral": 14}, "2026-07": {"recommend": 12, "neutral": 9, "warn": 2, "compare": 1}, "2026-08": {"recommend": 15, "neutral": 8, "warn": 1}, "2026-09": {"recommend": 6, "neutral": 3}, "2026-10": {"recommend": 1, "neutral": 2, "warn": 1}}, "quotes": [{"id": 802774880, "item_id": 50173430, "url": "https://news.ycombinator.com/item?id=50173430", "by": "apatheticonion", "date": "2026-10-20", "stance": "neutral", "quote": "I have repurposed my 6 year old 32gb M1 MacBook Pro to run Qwen 3.6 35b a3b", "excerpt": "Here's an anecdote. I have repurposed my 6 year old 32gb M1 MacBook Pro to run Qwen 3.6 35b a3b and I used to give it nightly maintenance tasks that span my OSS github repos. I gave it mundane tasks like updating dependencies, writing changelogs, etc. It ran about 2 hours a night. My MBP consumes 0.1w idle and about 10w running inference. I pay $0.33/kwh. If it runs for 2 hours a day over 365 days - it will cost $2.4", "context": "maintains OSS GitHub repos", "use_case": "nightly maintenance tasks on repos", "switched_from": "", "story": "DeepInfer: High-performance inference framework for DeepSeek models"}, {"id": 799810306, "item_id": 49988144, "url": "https://news.ycombinator.com/item?id=49988144", "by": "colingauvin", "date": "2026-10-07", "stance": "warn", "quote": "the coherence was really not great compared to frontier", "excerpt": "The author's configuration is..... Interesting. I don't disagree with the title. This spring folks were saying \"I use Qwen 3.6 27B (or 35 A3B) as a daily driver\". They either had low standards or were lying. The models were good enough technically, they capable enough, that you could see the potential, but the coherence was really not great compared to frontier. They'd lose track of what they were doing, mix things u", "context": "runs models on 2x Sparks", "use_case": "local daily driver model", "switched_from": "", "story": "Local AI is real now. And it's blowing my mind"}, {"id": 799494689, "item_id": 49968418, "url": "https://news.ycombinator.com/item?id=49968418", "by": "silentcuriousde", "date": "2026-10-05", "stance": "neutral", "quote": "trying to run models like Qwen3.6 35B A3B", "excerpt": "I recently open-sourced my coding agent CLI that I've been working on for the past while. It is a project I never intended to make public since it was part of my own personal local AI stack. But as I chipped away at it and made it actually usable as a daily driver, I thought it would be nice to make it public for others to see and use. I initially built it to see how much I could get the harness to make small models,", "context": "", "use_case": "local agentic coding", "switched_from": "", "story": "Reika \u2013 A coding agent CLI designed around small local models first"}, {"id": 798837842, "item_id": 49927365, "url": "https://news.ycombinator.com/item?id=49927365", "by": "razster", "date": "2026-10-01", "stance": "recommend", "quote": "Local model is Qwen3.6 35B-A3B and Qwen3.8 27B UD.", "excerpt": "I use my Pi to organize my graphic design assets as well as update documentation for client projects. It also is my daily driver for a story I'm working on and thanks to a ton of great skills created it has become more powerful. Local model is Qwen3.6 35B-A3B and Qwen3.8 27B UD. I love the compressing function, keeps on pushing. I've tried Oh-my-pi but it's too heavy for me and eats up 6% of my context on start. Also", "context": "", "use_case": "local model for Pi", "switched_from": "", "story": "Pi 1.0"}, {"id": 797939376, "item_id": 49871211, "url": "https://news.ycombinator.com/item?id=49871211", "by": "apatheticonion", "date": "2026-09-27", "stance": "neutral", "quote": "Qwen 3.6 a3b makes more mistakes than DeepSeek but it's free.", "excerpt": "Guide coding is more forgiving because the diffs are small enough that you can just try something and veto it if it's not good. Qwen 3.6 a3b makes more mistakes than DeepSeek but it's free. I'd imagine the next ~32gb-vram-class MoE model from Qwen will close the gap. The real deciding factor for me is inference speed. I use VSCode insiders with their BYO model configuration. I stage little bits at a time in a tight p", "context": "", "use_case": "local guided coding", "switched_from": "", "story": "GPT-6 Sol and Luna"}, {"id": 797400369, "item_id": 49837523, "url": "https://news.ycombinator.com/item?id=49837523", "by": "sznio", "date": "2026-09-24", "stance": "recommend", "quote": "it's both much faster and much more capable than Haiku", "excerpt": "Nothing, because current Haiku is useless. But I use Qwen3.6 35B-A3B and it's both much faster and much more capable than Haiku. And Qwen3.8 27B can do so much more, decently replaces Sonnet, albeit not as fast. Luna doesn't feel more capable nor faster than Qwen3.8. I think Anthropic could surprise us with some more innovation in this niche.", "context": "", "use_case": "", "switched_from": "Claude Haiku", "story": "Claude Opus 5.5"}, {"id": 797379840, "item_id": 49836240, "url": "https://news.ycombinator.com/item?id=49836240", "by": "LeBit", "date": "2026-09-24", "stance": "recommend", "quote": "I find them quite useful!", "excerpt": "I use the following models: - Qwen3.5 9B - Qwen3.6 35B A3B - Qwen3.8 27B - Gemma 4 26B A4B - Gemma 31B - Muse Glimmer 30B I have 48G. The MacMini is used solely for inference. llama.cpp + llama-swap. People are saying local models are crap and serves no purposes. I use them to help me spell check, write emails, write text messages, write JIRA tickets, write PR comments, etc. I also use them in coding agents to comple", "context": "Mac Mini with 48G", "use_case": "local inference for writing and coding", "switched_from": "", "story": "Best LLM for every budget, updated daily"}, {"id": 797043547, "item_id": 49815221, "url": "https://news.ycombinator.com/item?id=49815221", "by": "abalashov", "date": "2026-09-23", "stance": "neutral", "quote": "Qwen3.6-35B-A3B (speed) and Qwen3.8-27B (intelligence, but slow)", "excerpt": "Data points from my own usage, for whatever they're worth[1]. I'm just an individual contributor and hardly represent enterprise or large orgs: I've been on Kimi, with a little DeepSeek-V4-Pro, GLM 5.2/5.3, and MiMo thrown in, for probably about a year now. It's great here! 1) For DeepSeek, I recommend their Reasonix harness strongly, due to its alignment to DeepSeek's prefix cache. It means mostly (95%+) cache hit i", "context": "individual contributor", "use_case": "local experiments on MacBook", "switched_from": "", "story": "The current balance of power in open models"}, {"id": 796971186, "item_id": 49810699, "url": "https://news.ycombinator.com/item?id=49810699", "by": "apatheticonion", "date": "2026-09-23", "stance": "recommend", "quote": "running Qwen 3.6 35b a3b sipping 10w when generating tokens is good enough for a lot of my guide-coding work", "excerpt": "Can confirm. I have been using DeepSeek since forever and it's so good I was able to write a compiler and native desktop applications with it. I use it as a coding assistant in my IDE so the results end up at the same quality I would write by hand. I recently started a job that only uses Claude models. Opus and Sonnet are so slow you have no choice but to do multiple tasks in parallel. You create a git worktree, set ", "context": "", "use_case": "local guide-coding", "switched_from": "", "story": "GPT-6 Sol and Luna"}, {"id": 796953689, "item_id": 49809605, "url": "https://news.ycombinator.com/item?id=49809605", "by": "abalashov", "date": "2026-09-22", "stance": "neutral", "quote": "Qwen3.6-35B-A3B (speed)", "excerpt": "Welcome to the dark side. I've been on Kimi, with a little DeepSeek-V4-Pro, GLM 5.2/5.3, and MiMo thrown in, for probably about a year now. It's great here! For DeepSeek, I recommend their Reasonix harness strongly, due to its alignment to DeepSeek's prefix cache. It means mostly (95%+) cache hit input tokens, so very cheap large-scale code analyses and things that require mega context windows (at the cost of some at", "context": "MacBook with 128 GB unified memory", "use_case": "local model experiments", "switched_from": "", "story": "MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis"}, {"id": 794778483, "item_id": 49673655, "url": "https://news.ycombinator.com/item?id=49673655", "by": "seanmcdirmid", "date": "2026-09-12", "stance": "neutral", "quote": "I also use DeepSeek V4 flash and a Qwen 3.6 MoE for local stuff.", "excerpt": "Gemini 3.8 flash via Antigravity desktop has been my daily driver for the last few weeks. I also use DeepSeek V4 flash and a Qwen 3.6 MoE for local stuff.", "context": "", "use_case": "local", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 794771490, "item_id": 49673218, "url": "https://news.ycombinator.com/item?id=49673218", "by": "bellowsgulch", "date": "2026-09-12", "stance": "neutral", "quote": "honestly don\u2019t even bother using qwen3.6-35b-a3b now unless i need uncensored tasks finished, mostly reverse engineering", "excerpt": "mimo-v2.5-free, mimo-v2.5, deepseek-flash in that order, honestly don\u2019t even bother using qwen3.6-35b-a3b now unless i need uncensored tasks finished, mostly reverse engineering most engineering tasks don\u2019t require frontier llms when they get stuck, then i consider moving up to more capable models purchasing a claude plan seems widely unnecessary to me the tasks they do better than the average engineer cut both ways:", "context": "", "use_case": "uncensored reverse engineering tasks", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 792546001, "item_id": 49534125, "url": "https://news.ycombinator.com/item?id=49534125", "by": "ch_sm", "date": "2026-09-02", "stance": "recommend", "quote": "the smaller Gemma 4 models or qwen3.6:9b would probably run OK on that", "excerpt": "It depends on your use case, but the smaller Gemma 4 models or qwen3.6:9b would probably run OK on that. I recommend trying it, even just for fun. It\u2018s easy with omlx.", "context": "", "use_case": "local model on 24GB Mac", "switched_from": "", "story": "My local model setup on an M4 Pro Mac Mini"}, {"id": 792541394, "item_id": 49533837, "url": "https://news.ycombinator.com/item?id=49533837", "by": "jwr", "date": "2026-09-02", "stance": "recommend", "quote": "I run Qwen3.6-35B-A3B-MTP (fast!)", "excerpt": "For what it's worth, I've spent some time with Claude to develop a local runner for `llama.cpp`. I run Qwen3.6-35B-A3B-MTP (fast!) and Qwen3.8-27B (20 tokens/s). This was definitely worth the effort. Benchmarking and testing various approaches and various options really paid off. For example, one thing that surprised me was that MTP made things slower, not faster for Qwen3.8-27B. I use a 64GB MacBook Pro (M4 Max).", "context": "", "use_case": "local inference", "switched_from": "", "story": "Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/s"}, {"id": 792403488, "item_id": 49525218, "url": "https://news.ycombinator.com/item?id=49525218", "by": "prometheus1992", "date": "2026-09-01", "stance": "recommend", "quote": "I am running Qwen3.6-35B-A3B on my 16GB M3 and get 7-8 tokens/sec", "excerpt": "It's hard to believe 16GB unified memory will give you 5 tok/sec unless you are ignoring the thermal warnings. I am running Qwen3.6-35B-A3B on my 16GB M3 and get 7-8 tokens/sec with all the optimizations while keeping the peak memory and thermal warnings at check. https://github.com/deepanwadhwa/samosa-chat", "context": "", "use_case": "local LLM inference", "switched_from": "", "story": "Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/s"}, {"id": 792156881, "item_id": 49509805, "url": "https://news.ycombinator.com/item?id=49509805", "by": "gchamonlive", "date": "2026-08-31", "stance": "neutral", "quote": "I've used qwen3.6 and it's not as straightforward as", "excerpt": "I think 24gb is the bare minimum for a local qwen3.8 based setup. I've used qwen3.6 and it's not as straightforward as \"can it replace <insert the most cost-effective cloud solution today>\" Local llms don't suffer from cloud availability issues. Anyone that used Google models know that sometimes they just don't have capacity whatsoever, at least that was the state of things some months back when I used them. Just bea", "context": "", "use_case": "local LLM", "switched_from": "", "story": "Apple caught off guard by AI demand for Mac Mini and Mac Studio"}, {"id": 791704418, "item_id": 49481526, "url": "https://news.ycombinator.com/item?id=49481526", "by": "pwython", "date": "2026-08-28", "stance": "neutral", "quote": "definitely better than Qwen3.6 35b-a3b which was my go-to before", "excerpt": "Yea Qwen3.8 wasn't fun to use on my 64GB M4 Max either (better than these numbers though), so my new daily driver is Ornith-1.5-35B-A3B-MLX-4bit. I recommend giving that a whirl if you're on similar hardware, it's definitely better than Qwen3.6 35b-a3b which was my go-to before. https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B", "context": "", "use_case": "local LLM on Mac", "switched_from": "", "story": "Run Qwen3.8 27B locally: real numbers from my Mac Studio"}, {"id": 791579152, "item_id": 49473697, "url": "https://news.ycombinator.com/item?id=49473697", "by": "trucks-refinish", "date": "2026-08-28", "stance": "recommend", "quote": "It's a very capable model, not as smart as a frontier model but it is still very useful.", "excerpt": "I'm able to run qwen 3.6 35b a3b with my rtx 2060 super which also has 8gb of vram. It's a very capable model, not as smart as a frontier model but it is still very useful. I use an 6bit quant and get around 20 tokens per second.", "context": "runs on an RTX 2060 Super with 8GB VRAM", "use_case": "local LLM on 8GB GPU", "switched_from": "", "story": "Small Models Have Arrived"}, {"id": 791374224, "item_id": 49460889, "url": "https://news.ycombinator.com/item?id=49460889", "by": "Balinares", "date": "2026-08-27", "stance": "warn", "quote": "but those do exit eventually, unlike with Qwen 3.6.", "excerpt": "I've found it tends toward long thinking loops even for simple tasks (and any quantization seems to increase their length), but those do exit eventually, unlike with Qwen 3.6. I use the Unsloth UD_Q2_K_XL GGUF with default parameters, along with that custom template linked elsewhere in the thread, and no K/V cache quantization.", "context": "", "use_case": "local LLM", "switched_from": "", "story": "GLM-5.3-Flash"}, {"id": 791098032, "item_id": 49443627, "url": "https://news.ycombinator.com/item?id=49443627", "by": "cpburns2009", "date": "2026-08-26", "stance": "neutral", "quote": "I've only benchmarked 27B across Q3-6 quants using lm-eval.", "excerpt": "Not full precision. I've only benchmarked 27B across Q3-6 quants using lm-eval. I lack the hardware to bench 27B at BF16 but I might be able to do Q8_0. I haven't gotten around to doing 35B. I really should upload my collection of results to Github or somewhere. Here's a summary of what I have for 27B. I used unsloth's UD-Q{3-6}_K_XL quants across 11 evals. The values are pretty linear between Q3 and Q6. Qwen3.6-27B ", "context": "", "use_case": "benchmarking quantizations", "switched_from": "", "story": "Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)"}, {"id": 789377202, "item_id": 49336075, "url": "https://news.ycombinator.com/item?id=49336075", "by": "K0IN", "date": "2026-08-17", "stance": "neutral", "quote": "I used Qwen 3.6 27B extensively (>1B tokens)", "excerpt": "I used Qwen 3.6 27B extensively (>1B tokens) and DeepSeek V4 Flash (the older one also 2B+ tokens). And I just can't fathom that the new 3.8 beats the new DeepSeek V4 Flash (which, in my eyes, is one of the best everyday coding models). What an insane release, and convenient size to use every day/locally. but i will test this model extensivly.", "context": "", "use_case": "coding", "switched_from": "", "story": "Qwen3.8 27B scores 52 on Artificial Analysis"}, {"id": 789246176, "item_id": 49327886, "url": "https://news.ycombinator.com/item?id=49327886", "by": "digikata", "date": "2026-08-17", "stance": "recommend", "quote": "3.6 27B was easily a daily driver with only an occasional need to pop up to larger models for planning.", "excerpt": "Qwen3.6 27B is very usable, and dialing back thinking modes woth Qwen3.8 bring it close, but 3.8 stills feels slower. Unknown to me if the results are qualitatively better or worse overall - with the heavier thinking 3.8 felt worse in terms of coding tasks, but I think Im comparing a newly released model to one that has had a lot of harness tuning. 3.6 27B was easily a daily driver with only an occasional need to pop", "context": "", "use_case": "daily coding driver", "switched_from": "", "story": "Qwen 3.8 27B is excellent, but it defaults to overthinking things"}, {"id": 789240865, "item_id": 49327554, "url": "https://news.ycombinator.com/item?id=49327554", "by": "hellajack3d", "date": "2026-08-17", "stance": "neutral", "quote": "This was mainly put together to rein in Qwen3.6-27B", "excerpt": "I forked llama.cpp and added some crude mechanism to keep exactly this behavior under control - essentially guiding the reasoning process by injecting text strategically at specific thresholds. This was mainly put together to rein in Qwen3.6-27B, but I'd imagine 3.8 would react similarly. Fork can be found here - https://github.com/laurencehardman/llama-mindcontrol/tree/master Of course hacks like this are not perfec", "context": "", "use_case": "local model", "switched_from": "", "story": "Qwen 3.8 27B is excellent, but it defaults to overthinking things"}, {"id": 789218880, "item_id": 49326180, "url": "https://news.ycombinator.com/item?id=49326180", "by": "hedgehog", "date": "2026-08-16", "stance": "recommend", "quote": "I use Qwen 3.6 a fair amount using his template and it fixes some issues I saw with the upstream versions.", "excerpt": "Look for improved templates by \"froggeric\" on Hugging Face. I use Qwen 3.6 a fair amount using his template and it fixes some issues I saw with the upstream versions.", "context": "", "use_case": "local model", "switched_from": "", "story": "Qwen 3.8 27B is excellent, but it defaults to overthinking things"}, {"id": 788811152, "item_id": 49300697, "url": "https://news.ycombinator.com/item?id=49300697", "by": "lrvick", "date": "2026-08-14", "stance": "recommend", "quote": "Use 3.6 27b as a daily driver for months with charmbracelet crush.", "excerpt": "Use 3.6 27b as a daily driver for months with charmbracelet crush. Gemma 26b-A3b is not even remotely comparable in terms of coding for me. YMMV depending on how you work, what harness you use, etc I suppose.", "context": "", "use_case": "daily coding", "switched_from": "", "story": "Qwen 3.8 27B is out: open weights, best local dense model yet"}, {"id": 788644962, "item_id": 49290310, "url": "https://news.ycombinator.com/item?id=49290310", "by": "SwellJoe", "date": "2026-08-13", "stance": "recommend", "quote": "the best models you can run at home are probably Qwen 3.6 27B and Gemma 4 31B", "excerpt": "\"used to\" in your sentence already answers your curiosity. ROCm runs most things just fine without any ceremony or difficulty. I have a couple of the same cards as covered in the article, bought before they got expensive (I'd recommend current Radeon AI Pro 9700 cards over the V620 now, though, as they have increased in price since I bought mine), and they're at the long end of the supported chart for ROCm...nearly E", "context": "", "use_case": "local models", "switched_from": "", "story": "We Have AI at Home Chapter 1: A Box of Scraps"}, {"id": 788038720, "item_id": 49252420, "url": "https://news.ycombinator.com/item?id=49252420", "by": "qudat", "date": "2026-08-10", "stance": "recommend", "quote": "I use qwen3.6 27b for personal projects and I think it works pretty great.", "excerpt": "> Or maybe models get so good that a 30B model is genuinely smart enough to do everything, so nobody really needs a model like Opus or Sol unless they\u2019re trying to solve the Reimann Hypothesis. I don\u2019t really buy this. Models can do frontier mathematical work today while still being not smart enough to refactor large codebases as well as me, so it\u2019s hard to imagine a world where I don\u2019t just want to use the smartest ", "context": "", "use_case": "personal projects, local model", "switched_from": "", "story": "No, local models will not win"}, {"id": 787992352, "item_id": 49249522, "url": "https://news.ycombinator.com/item?id=49249522", "by": "bitexploder", "date": "2026-08-10", "stance": "recommend", "quote": "My final 35B A3B tip -- use an extremely simple harness.", "excerpt": "I have a custom A3B proxy that caps its thinking off. It is a known issue with the model that Qwen themselves documented but is almost never addressed in any harnesses. I also patched up a few other known bugs in the proxy. I highly recommend you shim A3B and when it hits 2K thinking tokens inject (paraphrasing) 'Time to wrap it up bud! Get to work' into its thinking stream. It almost always gets to work. If it needs", "context": "runs a custom proxy for local models", "use_case": "local agentic coding", "switched_from": "", "story": "Meta Muse Glimmer \u2013 open weights 30B local coding model"}, {"id": 787977649, "item_id": 49248603, "url": "https://news.ycombinator.com/item?id=49248603", "by": "dofm", "date": "2026-08-10", "stance": "recommend", "quote": "I love the 3.6 version", "excerpt": "One thing I would caution is staying out of the prediction market like this. Tech tends to get boring when you judge current products against the hypothetical capabilities of unannounced products that may never ship. It's like comparing Nikon cameras against Canon camera rumours, or comparing iPhones against unannounced and therefore largely imaginary Samsungs. - If they do a Qwen 3.8 35B A3B (and I hope they do beca", "context": "", "use_case": "local models", "switched_from": "", "story": "Meta Muse Glimmer \u2013 open weights 30B local coding model"}, {"id": 787905568, "item_id": 49244098, "url": "https://news.ycombinator.com/item?id=49244098", "by": "Roark66", "date": "2026-08-10", "stance": "recommend", "quote": "The moe works well in Librechat combined with my local search/Web retrieval system.", "excerpt": "I've been using Qwen3.6 models locally for a couple of weeks. Both the A3B moe and the dense variant. The moe works well in Librechat combined with my local search/Web retrieval system. All components use open source projects such as SearXNG, Crawl4AI, MetaMCP, Jina rerank, but all needed quite a bit of coding to work nicely together. I get 140 tok/s on short prompts on an rtx3090 on the qwen3.6 moe which makes is ea", "context": "", "use_case": "local web research and planning", "switched_from": "", "story": "DeepSeek costs OpenCode Go user $1.14/day; dual DGX breaks even in 24 years"}, {"id": 787638480, "item_id": 49227405, "url": "https://news.ycombinator.com/item?id=49227405", "by": "h4fizwasabie", "date": "2026-08-08", "stance": "neutral", "quote": "i used to play around with qwen 3.6 35b a3b with 17/tps", "excerpt": "i need some experience from anyone. which model is good for local usage. i prefer moe models, since im only running 4gb vram. i used to play around with qwen 3.6 35b a3b with 17/tps. its been few months since i last play around with local LLM. is there any improvement on local ai development?", "context": "runs 4GB VRAM", "use_case": "local LLM on 4GB VRAM", "switched_from": "", "story": "U.S. Department of Energy Launches the Genesis Open Models Initiative"}, {"id": 787438721, "item_id": 49214920, "url": "https://news.ycombinator.com/item?id=49214920", "by": "nozzlegear", "date": "2026-08-07", "stance": "recommend", "quote": "Then I have Qwen3.6-35B-A3B as the \"orchestrator\" agent (it's just what I use for my daily driver)", "excerpt": "Yeah, they're excellent. I use a small Qwen3-4B-Instruct as the \"actor\" agent which goes out to the website, searches for a batch of items, decides which best matches the criteria, and puts them in the cart. Then I have Qwen3.6-35B-A3B as the \"orchestrator\" agent (it's just what I use for my daily driver), it's responsible for chatting with my wife about the menu, recipes and ingredients, and tells the smaller agent ", "context": "", "use_case": "orchestrating shopping agent", "switched_from": "", "story": "Kitesurf: Agent-first browser that runs in V8 isolates"}, {"id": 787288387, "item_id": 49205524, "url": "https://news.ycombinator.com/item?id=49205524", "by": "alasr", "date": "2026-08-06", "stance": "recommend", "quote": "start with Qwen3.6, Google's Gemma4, OpenAI's gpt-oss-20b or Nvidia's Nemotron", "excerpt": "- See \"Llama.cpp vs Ollama\"[0] and \"Llama.cpp vs LM Studio\"[1] for a high-level features comparison and find out which one suite you best (check out the \"Target Users\" category[0][1], at least). - If you're still not sure, \"LM Studio\"[2] is not bad to start with as you'll be able to download, start/stop/manage and chat with your LLM model all in one place: a single desktop app. Also, once installed, enable \"Developer", "context": "", "use_case": "local LLM", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786833233, "item_id": 49177077, "url": "https://news.ycombinator.com/item?id=49177077", "by": "megadragon9", "date": "2026-08-04", "stance": "neutral", "quote": "local Qwen3.6-35B A3B", "excerpt": "I did something similar. It started off as a \"self-improving agent\" project, inspired by autoresearch, then later on I reframed it as \"harness training\" (discrete program search) borrowing the mental model from ML training. I \"trained\" the harness on a subset of Terminal-Bench 2.0 tasks while keeping the LLM (local Qwen3.6-35B A3B) frozen. Making LLM inference and the task environments fully deterministic was necessa", "context": "", "use_case": "frozen local LLM for harness training", "switched_from": "", "story": "Harness Engineering for Self-Improvement"}, {"id": 786523904, "item_id": 49157744, "url": "https://news.ycombinator.com/item?id=49157744", "by": "jkingsman", "date": "2026-08-03", "stance": "recommend", "quote": "I use it for personal tasks that are just not worth the claude tokens", "excerpt": "Not OP, but I use it for personal tasks that are just not worth the claude tokens -- rooting through historical medical records to unify prescription history, super-OCR'ing thousands of PDF pages (i.e. beyond PDF dumping -- vision means it can look at tables, understand tricky things like a continuation of a block quote or aside on the next page, etc.), and when power is cheap I'll just let it noodle on little projec", "context": "runs models locally", "use_case": "local OCR and personal data tasks", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786510561, "item_id": 49156910, "url": "https://news.ycombinator.com/item?id=49156910", "by": "AbsurdCensor", "date": "2026-08-03", "stance": "recommend", "quote": "Sure a lot of this could be done without AI, but it's certainly quicker and easier", "excerpt": "Not OP, but I use it for a ton of smaller things. I have it hooked into Hermes and have been using it to help bulk rename my media folders so they all follow a common format, add titles that sort of thing which wouldn't be easy to 'script'. Another thing I use it for is comparing data sets, looking at my exported Spotify artists and compare to what I have locally, and letting me know where there are missing artists, ", "context": "runs local AI box on solar", "use_case": "local bulk non-code tasks", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786461648, "item_id": 49153853, "url": "https://news.ycombinator.com/item?id=49153853", "by": "harhargange", "date": "2026-08-03", "stance": "neutral", "quote": "I used the qwen 3.6 models and convinced myself to go ahead and get a gpu.", "excerpt": "I used the qwen 3.6 models and convinced myself to go ahead and get a gpu. I got the 7900 xt with 20 gb vram.", "context": "", "use_case": "local models", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786457408, "item_id": 49153588, "url": "https://news.ycombinator.com/item?id=49153588", "by": "cybertim", "date": "2026-08-03", "stance": "recommend", "quote": "I also \"evolved\" into 27b (q8 unsloth)", "excerpt": "I also \"evolved\" into 27b (q8 unsloth) and pi.dev (tried many combinations) feels for me the same as opus 4.5 that i use at work, faster even (using 2x 3080 20GB gives me 60-80tk/s). Though you do need to feed it more details up front (about what exactly you are planning to do and a good written skill.md) but I work that way anyways, im hyped for 3.8", "context": "", "use_case": "local coding model", "switched_from": "", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786414864, "item_id": 49150929, "url": "https://news.ycombinator.com/item?id=49150929", "by": "nozzlegear", "date": "2026-08-02", "stance": "recommend", "quote": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April.", "excerpt": "Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April. The Qwen3.6 line is easily the best local model I've tried, and I've tried a lot. I've got it diligently grinding away on my laptop right now, reviewing and fixing some bugs in my F# code.", "context": "", "use_case": "local coding and bug fixing", "switched_from": "Claude", "story": "Qwen3.8-Max: A New Bar for Coding and Cowork"}, {"id": 786356289, "item_id": 49147268, "url": "https://news.ycombinator.com/item?id=49147268", "by": "redrove", "date": "2026-08-02", "stance": "recommend", "quote": "I have qwen 3.6 27b in an rtx 6000 pro as well so I use that as a workhorse in pi", "excerpt": "The FP8 version from DeepSeek themselves [0], around 1800 tps prefill and 45 tokens per second decode. I\u2019ve been running a custom VLLM image with b12x as well as nvfp4_ds_mla. I would say it\u2019s quite fantastic in day to day, I use it mostly in Hermes and sometimes for coding. I have qwen 3.6 27b in an rtx 6000 pro as well so I use that as a workhorse in pi with DS as a reviewer/planner. [0] https://huggingface.co/deep", "context": "", "use_case": "workhorse coding model", "switched_from": "", "story": "Show HN: NixOS-DGX-Spark \u2013 Nix and NixOS on the DGX Spark"}, {"id": 785947718, "item_id": 49121732, "url": "https://news.ycombinator.com/item?id=49121732", "by": "user43928", "date": "2026-07-31", "stance": "warn", "quote": "\"I love hallucinated garbage\": Sonnet, Qwen 3.6, GPT 5.4 mini, GPT 5.3 Codex", "excerpt": "In my mind there are three tiers: The SOTA: Fable, GPT 5.6 Sol, Opus 5 The \"enterprise admin did not turn on the new models\": Opus 4.8, GPT 5.5 The \"I love hallucinated garbage\": Sonnet, Qwen 3.6, GPT 5.4 mini, GPT 5.3 Codex, etc. Results vary widely", "context": "", "use_case": "", "switched_from": "", "story": "Google fixed more Chrome bugs in June than over the past two years, thanks to AI"}, {"id": 783464064, "item_id": 48966504, "url": "https://news.ycombinator.com/item?id=48966504", "by": "nsbk", "date": "2026-07-19", "stance": "recommend", "quote": "a single 3090 already provides a great experience at a reasonable quant and context size", "excerpt": "I run a 2x 3090 rig, but a single 3090 already provides a great experience at a reasonable quant and context size. On a single card I used to run Qwen_Qwen3.6-27B-Q4_K_M or similarly quantized 35B MoE at 65536 context size", "context": "runs 2x 3090 rig", "use_case": "local inference on 3090", "switched_from": "", "story": "Qwen3.8 is launching and going open-weight soon"}, {"id": 782152256, "item_id": 48884516, "url": "https://news.ycombinator.com/item?id=48884516", "by": "docheinestages", "date": "2026-07-12", "stance": "warn", "quote": "I tried Qwen 3.6 27B, but I can only get around 10 tokens per second, but it's painfully slow, and it often fails during `write_file` tool calls", "excerpt": "I'm on a 48 GB M5 Macbook Pro. I use 4-bit quants with a context window of 16-32k. I tried Qwen 3.6 27B, but I can only get around 10 tokens per second, but it's painfully slow, and it often fails during `write_file` tool calls, even with Qwen Code.", "context": "48 GB M5 Macbook Pro user", "use_case": "local coding model", "switched_from": "", "story": "Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k"}, {"id": 779714784, "item_id": 48732174, "url": "https://news.ycombinator.com/item?id=48732174", "by": "ThunderSizzle", "date": "2026-06-30", "stance": "recommend", "quote": "35B is pretty good", "excerpt": "I'll give 27B-MTP a try. I think I can tolerate 45 tps if the results are technically better. 35B is pretty good, but definitely shows it's inabilities at times (probably either due to the heavy caching quantization I'm doing, or the heavy model quantization vs what 2 GPUs could run). My biggest gripe is that both pi and opencode seem to have trouble parsing the thinking blocks at times, and the model sometimes cuts-", "context": "runs local models on 2 GPUs", "use_case": "local coding model", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 779670192, "item_id": 48729387, "url": "https://news.ycombinator.com/item?id=48729387", "by": "PeterStuer", "date": "2026-06-29", "stance": "warn", "quote": "for software development it is unsurprisingly a dramatic downgrade from the Opus I use as my daily driver.", "excerpt": "Been running it on a 9950x3D with 96GB and a 4090. Speedwise it is fine. Bit while not completely useless, for software development it is unsurprisingly a dramatic downgrade from the Opus I use as my daily driver.", "context": "", "use_case": "local software development", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 779649472, "item_id": 48728092, "url": "https://news.ycombinator.com/item?id=48728092", "by": "m3kw9", "date": "2026-06-29", "stance": "warn", "quote": "it can't get past a simple coding test that 5.5 passes with light reasoning", "excerpt": "Hmm, i used it and it can't get past a simple coding test that 5.5 passes with light reasoning", "context": "", "use_case": "coding", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 779606528, "item_id": 48725408, "url": "https://news.ycombinator.com/item?id=48725408", "by": "cpburns2009", "date": "2026-06-29", "stance": "recommend", "quote": "A 32gb card does run it nicely.", "excerpt": "A 32gb card does run it nicely. I use unsloth's UD-Q5_K_XL at 256k context (k/v at q8_0), and get ~67 t/s on a 5090. I still need to look into MTP.", "context": "", "use_case": "local LLM on a 5090", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 779581937, "item_id": 48723871, "url": "https://news.ycombinator.com/item?id=48723871", "by": "Zambyte", "date": "2026-06-29", "stance": "recommend", "quote": "I have been very blown away by how well it works.", "excerpt": "I have been using pi (and previously the codex cli) with Qwen 3.6 27b with 100k context for my development at work, and I have been very blown away by how well it works. It's not perfect, but it's enough to accelerate my normal development flow. I mostly use it for writing Go and C#.", "context": "", "use_case": "local coding with Go and C#", "switched_from": "", "story": "Qwen 3.6 27B is the sweet spot for local development"}, {"id": 778616800, "item_id": 48663550, "url": "https://news.ycombinator.com/item?id=48663550", "by": "crazycracker", "date": "2026-06-24", "stance": "recommend", "quote": "I use qwen3.6:27b as a daily driver and I am heavily impressed with it.", "excerpt": "Historically the cost of compute has also gone down. Like just look at it as compared to a year ago. We have amazing open source models that can run on consumer hardware and if we go away from our obsession of using opus 4.8 or mythos for everything then it actually is super amazing to see what these open source models could do. I use qwen3.6:27b as a daily driver and I am heavily impressed with it.", "context": "", "use_case": "daily local model", "switched_from": "", "story": "For Most of the World, Open-Source AI Is the Only Way Forward"}]}