{"slug": "open-source-models-for-coding", "title": "Open source models for coding?", "summary": "A developer seeking open source coding LLMs for a 128GB Ryzen AI Max system reports that Gemma 4 and Qwen 3.6 underperform, while community members recommend Qwen 3.5 35B and 122B, DeepSeek V4 Flash, Mistral 3.5 Medium, and Step 3.7, noting that proper quantization and settings are critical. Qwen announced that 3.8 open weights models will be released soon, with hopes for a ~100B MoE model. Unsloth Studio is highlighted as a tool for running and training models locally.", "body_md": "I am looking for good open source LLM options that can be used for coding in something like opencode. I tried Gemma 4 and Qwen 3.6 and find them to be underwhelming and perform poorly (in terms of output quality, not in terms of execution speed).\n\nI am trying to evaluate options that can be run on a 128GB Ryzen AI Max system and intend to purchase hardware eventually, but until then I will be trying out using cloud providers.\n\nWhat models should I look at?\n\n1 Like\n\nReality check: you’re not going to be one-shotting apps on local LLMs like you would in the cloud. That’s not the use case they’re designed for; doing actual work through an agent is a very different experience with them.\n\nRun by an agent, like Cline or OpenCode (for example), Qwen 3.6 35B or 27B can be excellent - if configured correctly. That means setting up the full context, full GPU offload etc.\n\nI use 35B on a 1.2 million line Ruby codebase with Cline, and not only does it work well, it often spots stuff that Opus 4.7 misses (that’s not to say that it’s *better* than Opus 4.7 in general, mind). With my R9700, it’s also noticeably faster and recovers better from tool call failures.\n\nSo…the question is: what are you planning to use it for, and how are you planning to use it?\n\n2 Likes\n\nI have these small models often failing after multiple iterations for simple tasks as “rewrite this 30 line batch script into bash” while claude sonnet can one shot it without problems.\n\n*Which* models? And which quants?\n\nWhat models are you using that are failing? Qwen3.6.-27B definitely shouldn’t be failing, if you are using some low tier quant; definitely check out Unsloth because even UD_Q4_K_XL shouldn’t fail, but you need more than 16GB VRAM… 24GB minimum.\n\nI Have heard that Qwen3.5-9b is decent, and some people have made 3.6-9B now but you need again to do UD_Q8_K_XL if you really want the most from it.\n\n1 Like\n\nOne of my systems is a 128gb strix halo, and there are some great coding models that can run on it. First for the models you were having programs with.\n\n- Make sure you have a good quant, don’t go lower then 4 but 8 is better.\n- Inference and loading settings. On a stix halo always turn mmap off. For inference settings for a model like qwen 3.5 35b or the 122b that a strix halo can run in q6 with full context length I set temp to .65, top k 20, repeat 1.03, top p .95, bottom p .03. You can play with the number of active experts aswell I run the 122b with 20 active.\n\nFor running models on strix halo you should definitely look at these strix halo toolboxs [Strix Halo AI Toolboxes](https://strix-halo-toolboxes.com/#about)\n\nFor running models on any system everyone should check out unsloth studio. You can run and train models in it. There is auto model loading and inference settings if you want, tool calls correction, speed increase and more. [Introducing Unsloth Studio | Unsloth Documentation](https://unsloth.ai/docs/new/studio)\n\nModels I’m running qwen 3.5 122b or three of the 3.6, deepseek v4 flash, mistral 3.5 medium (it is slow but great I may make a MOE version to fix the speed), step 3.7. Let me know if you need more.\n\n2 Likes\n\nQwen just announced that 3.8 will be released soon, and there will be 3.8 open weights models.\n\n3.6 didn’t have a large MOE model, and 3.7 didn’t have open weights, so there are many people hoping 3.8 will have a ~100B MOE model.\n\n1 Like\n\nI’m actually doing a lot of testing for them and putting in tons of bug reports continuously. And together we are improving the platform rather quickly. I’ve been basically using it every single day. So almost every bug I am finding it and writing detailed reports on it, or at least as detailed as it needs to be. I definitely recommend that anybody going into local AI, check out Unsloth Studio because I think that it is the best mixture of everything that people are likely to use. From a local LLM outside of a direct agent for programming or some other specific task.\n\nThe software also does great for training, which I am not currently into, but have only done testing for, and does this well on AMD, even in VRAM limited scenarios.\n\nWell, the 3.5 had the 122 A10B version, but honestly, that one is actually worse than the 27B model when it comes to deep thinking.\n\nThe most dense of the entire QWEN family was actually the 27B. Because even the 397B only has 22B active.\n\nI would welcome a 50-80B dense like what we got with Qwen3.\n\n1 Like\n\nPoint of order - there was no 50-80B dense model in the Qwen3 family, as far as I know. There was Qwen3-Coder-Next 80B, but that was MoE with 3B active.\n\nQwen3 Coder 80B was MoE? My mistake then, I thought it was Dense.\n\nWith only 3B active, no wonder it wasn’t very good when I tried it… I had the best luck with Qwen3-14B-BF16 originally.\n\nThe Qwen3.6 27B dense model while slower than the 35B MoE, is one of the top models that you can run on a machine with 128 GB (and 64 GB with Q8).\n\nI use both of them depending on the task, if it is coding or architecture discussions. My coding is usually long horizon agentic coding tasks, using opencode.\n\nTry BF16 as test, it “should” at least solve your tasks.\n\n1 Like\n\nI like to use the Unsloth Dynamic Quantization versions in Q8 XL tensor size. So essentially it has BF16 tensors for the important ones and quantizes ones that are less important.", "url": "https://wpnews.pro/news/open-source-models-for-coding", "canonical_source": "https://forum.level1techs.com/t/open-source-models-for-coding/251758#post_13", "published_at": "2026-08-27 07:20:44+00:00", "updated_at": "2026-08-27 07:49:54.258789+00:00", "lang": "en", "topics": ["large-language-models", "ai-tools", "developer-tools"], "entities": ["Gemma 4", "Qwen 3.6", "Qwen 3.5", "DeepSeek V4 Flash", "Mistral 3.5 Medium", "Step 3.7", "Unsloth Studio", "Ryzen AI Max"], "alternates": {"html": "https://wpnews.pro/news/open-source-models-for-coding", "markdown": "https://wpnews.pro/news/open-source-models-for-coding.md", "text": "https://wpnews.pro/news/open-source-models-for-coding.txt", "jsonld": "https://wpnews.pro/news/open-source-models-for-coding.jsonld"}}