Qwen Coder
Alibaba's family of open-weight code-specialized large language models from the Qwen series.
HN commenters like it as a local, private coding model1234 that runs on consumer hardware and works with agents like OpenCode.5678 The most common caveat is that it trails cloud frontier models such as Claude and Codex,91011 and some report bad code from small variants or unreliable tool calling.1213
Why people recommend it
Local and private · 3
Commenters value running it locally, sometimes air-gapped, so code and data stay in-house.
“if you want something local and private Qwen 2.5 coder has been good at this time.”
“it worked surprisingly well”
Show 1 more
“currently Qwen3.5 122B A10B and Qwen3 Coder Next”
Good results · 3
Users report strong coding output for a local model, including best results among similar small models.
“So far Qwen3-coder gives the me best results.”
“Qwen3-Coder-Next also remains amazing as a local model.”
Show 1 more
“This was with a 30b model at 4-bit quantization, so pretty tiny by LLM standards.”
Consumer hardware · 3
It runs on laptops, single GPUs and even a 64 GB RAM machine without a GPU.
“I run Gemma 4 26b (mixture of experts) and Qwen 3 coder with Ollama.”
“So far Qwen3-coder gives the me best results.”
Show 1 more
“This was with a 30b model at 4-bit quantization, so pretty tiny by LLM standards.”
Agentic coding · 3
It is used with OpenCode and other agents for local agentic coding sessions.
“I think it might actually be the best option for agentic coding on the 395+ for now.”
“This was with a 30b model at 4-bit quantization, so pretty tiny by LLM standards.”
Show 1 more
“Qwen3-Coder-Next-GGUF:UD-Q4_K_XL to run local agentic coding sessions”
Speed · 3
Commenters describe decent speed, with prompts often returning in under 30 seconds.
“Qwen3 coder is pretty good and decent speed.”
“I run Qwen3-Coder-Next-UD-Q4_K_XL and other than the initial wait to initialise context”
Show 1 more
“A 3B resident parameter MOE allows absolutely huge savings on inference costs.”
Caveats and dissent
Below frontier · 3
Local runs do not match cloud frontier models like Claude Sonnet or Codex.
“various local models like Qwen Coder Next to work on various tasks”
“So far Qwen3-coder gives the me best results.”
Show 1 more
“if you want something local and private Qwen 2.5 coder has been good at this time.”
Small-model quality · 1
One user found Qwen2.5-Coder on a 2 GB GPU generated bad Python and JavaScript.
“However this model generates bad code”
Tool calling · 1
One user had to remind it constantly to invoke its tools under continue.dev.
Does Qwen3 Coder do a good job invoking its tools as appropriate for you? Under continue.dev at least, I've found I need to remind it constantly.
Setup tuning · 2
Getting GPU use or higher speeds required fiddling with inference settings.
“I run Qwen3-Coder-Next-UD-Q4_K_XL and other than the initial wait to initialise context”
“custom Burn inference for Qwen3-Coder-Next”
Hosted rate limits · 1
On a hosted subscription, tools quickly hit per-minute request limits with the 480B model.
“none of the “supported” tools I tried were able to use it in a way that didn’t hit the per minute request limit in a few seconds.”
Often compared with
Who recommends it in their own words
Threads where it came up
- Ask HN: Who uses open LLMs and coding assistants locally? Share setup and laptop 2 mentions
- OpenShell – A safe, private runtime for fleets of autonomous AI agents 1 mention
- Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware 1 mention
- Ask HN: How is everyone using Local LLMs? 1 mention
- New Framework Desktop Option with AMD Ryzen AI Max+ Pro 495 and 192GB Memory 1 mention
- ZCode: Claude Code from the Makers of GLM 1 mention
Also written as Qwen Coder, Qwen2.5-Coder, Qwen3 Coder, Qwen3-Coder. Data: qwen-coder.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 23
“it worked surprisingly well”
“However this model generates bad code”
“various local models like Qwen Coder Next to work on various tasks”
“I run Qwen3-Coder-Next-UD-Q4_K_XL and other than the initial wait to initialise context”
“Qwen3-Coder-Next-GGUF:UD-Q4_K_XL to run local agentic coding sessions”
“This was with a 30b model at 4-bit quantization, so pretty tiny by LLM standards.”
“I run Gemma 4 26b (mixture of experts) and Qwen 3 coder with Ollama.”
“custom Burn inference for Qwen3-Coder-Next”
“if you want something local and private Qwen 2.5 coder has been good at this time.”
“currently Qwen3.5 122B A10B and Qwen3 Coder Next”
“Qwen3-Coder-Next also remains amazing as a local model.”
“Qwen2.5-Coder 32B is genuinely impressive for local inference”
“I think it might actually be the best option for agentic coding on the 395+ for now.”
“You can run smaller models locally that are pretty good at code gen.”
“Qwen3 coder is pretty good and decent speed.”
“The smaller Qwen3 models (like qwen3-coder) I use in tandem”
“A 3B resident parameter MOE allows absolutely huge savings on inference costs.”
“So far Qwen3-coder gives the me best results.”
“I've got a Linux chatbot helper on the "test this" pile with Qwen Coder - not evaluated it”
“a custom agent for Qwen Coder via Llama.cpp [1]”
“none of the “supported” tools I tried were able to use it in a way that didn’t hit the per minute request limit in a few seconds.”
“mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit”
Does Qwen3 Coder do a good job invoking its tools as appropriate for you? Under continue.dev at least, I've found I need to remind it constantly.