# Qwen3-Coder 165 Hacker News commenters recommended Qwen3-Coder and 49 criticized it, November 2025 to November 2026. It was recommended in 12 of 13 months; owners who said how long report a median of 0.17 years (7 owners). It meets the Hacker Approved badge rule. **Verdict:** A free open-weight coding model that runs locally at usable speed and handles routine coding and tool calling well, though it trails Claude on harder work and can loop. ## Why people recommend it - Best coding model to run locally (6 comments): "it's actually seeming pretty competent. These are the first results I've had on a locally hosted model that I could see myself actually using" ([source](https://news.ycombinator.com/item?id=46875833)) "qwen 3/3.6 coder is the real champion doing almost the same things with less than 1/10th the memory requirements" ([source](https://news.ycombinator.com/item?id=48407646)) "If you can fire up quantized Qwen 2.5 coder model on your own machine, you control your workflow." ([source](https://news.ycombinator.com/item?id=49757329)) - Close to Claude for everyday coding (5 comments): "I was impressed at how good Qwen3-Coder is. Not as good as Claude, but close - maybe even better in some ways." ([source](https://news.ycombinator.com/item?id=45835406)) "I switched from Claude to Qwen3-Coder-Next for day-to-day on dev projects and don't see a big difference." ([source](https://news.ycombinator.com/item?id=46893770)) "We tend to find Qwen3-Coder-Next better at coding at least on our anecdotal examples from our codebases." ([source](https://news.ycombinator.com/item?id=47201814)) - Fast, efficient MoE on modest hardware (5 comments): "GPU for prompt processing, ram for model weights and context length, and it being MoE makes it fairly zippy." ([source](https://news.ycombinator.com/item?id=46876420)) "Worked great and was so stupidly fast it made me realize that if the hardware can be at that level" ([source](https://news.ycombinator.com/item?id=48213642)) "it gets 40-50 Token/s with 65k context. That is a lot more usable than a Quantised Qwen3.8-27B" ([source](https://news.ycombinator.com/item?id=50129753)) - Reliable tool calling and instruction following (3 comments): "accuracy around function/tool calling hasn't been a problem for me now for about 6 - 12 months, personally, since around QwenCoder3" ([source](https://news.ycombinator.com/item?id=48204556)) "Qwen3 is really good at hitting the tools first, then coming up with a well-grounded answer based on reality." ([source](https://news.ycombinator.com/item?id=45828875)) "Qwen3-Code is way more likely to do exactly what is being asked than any Gemini model" ([source](https://news.ycombinator.com/item?id=46900750)) ## What people warn about - Gets stuck in tool-call loops (3 comments): "I've also seen the occasional tool call looping in Qwen3 Coder Next, that seems to be an easy failure mode for that model to hit." ([source](https://news.ycombinator.com/item?id=47066027)) "Got caught in a loop reading a single file over and over again until the context filled up." ([source](https://news.ycombinator.com/item?id=46887710)) "when an open model (qwen3-coder:30b) drives an mcp tool (chrome-devtools-mcp@latest), it hallucinates the tool calls" ([source](https://news.ycombinator.com/item?id=49936051)) - Weaker than Claude on serious coding (5 comments): "Qwen3 variants perform very well, speed wise. For local sensitive document work, these are excellent; for serious coding not so much." ([source](https://news.ycombinator.com/item?id=46892439)) "at least the most recent versions are getting better but still not quite good enough in real world use cases outside of benchmarks" ([source](https://news.ycombinator.com/item?id=49359045)) "it's really not in the same universe with Claude Code, it will be slow and generate bad code." ([source](https://news.ycombinator.com/item?id=46912510)) - Slow loading and generation locally (4 comments): "I get 6-7 tok/s, which is slow for throwing around program code, but fine for general knowledge queries." ([source](https://news.ycombinator.com/item?id=48498901)) "QWEN3.5-Next-Coder does wonders. It's drawbacks are time to first token is 30 seconds to load the model" ([source](https://news.ycombinator.com/item?id=47662654)) "Any of coder-next quants larger than UD-Q4_K_XL take forever to load, especially with ROCm." ([source](https://news.ycombinator.com/item?id=47620275)) - Wastes time with wrong output (3 comments): "I tried and it would just bug out, things like respond with JSON strings summarizing what I asked of it or just outright getting things wrong entirely" ([source](https://news.ycombinator.com/item?id=46880209)) "Doing anything beyond basic work will just burn tokens for hours while you review and reject code." ([source](https://news.ycombinator.com/item?id=46976164)) "both can get some things done and can waste tons of time getting some things completely wrong" ([source](https://news.ycombinator.com/item?id=47814223)) ## Alternatives hackers recommend - [Cursor Composer](https://hackerapproved.com/p/composer-2-5.md): 214 recommend - [GPT-5.3-Codex](https://hackerapproved.com/p/gpt-5-3-codex.md): 135 recommend - [Meta Muse Glimmer 30B](https://hackerapproved.com/p/meta-muse-glimmer.md): 70 recommend - [GPT-5.2-Codex](https://hackerapproved.com/p/gpt-5-2-codex.md): 54 recommend - [Laguna S 2.1](https://hackerapproved.com/p/laguna-s-2-1.md): 43 recommend ## Cite as Hacker Approved, "Qwen3-Coder", https://hackerapproved.com/p/qwen3-coder-next/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).