# Qwen 3.8 27B 1173 Hacker News commenters recommended Qwen 3.8 27B and 426 criticized it, November 2025 to November 2026. It was recommended in 13 of 13 months; owners who said how long report a median of 0.125 years (32 owners). It meets the Hacker Approved badge rule. **Verdict:** An open-weights model that runs on a single high-end GPU and nears frontier coding quality, though it overthinks, burns tokens and struggles with complex tasks. ## Why people recommend it - Near-frontier coding quality (9 comments): "it FEELS like opus 4.5, i gave it an image and a broad overview of what I wanted it to build, and it built the whole thing from beginning to end." ([source](https://news.ycombinator.com/item?id=49303069)) "Many of them (K2.5 the Minimax, the GLM models, the Qwen 3.6 models) are about as capable as frontier US models from about 4 months ago or so." ([source](https://news.ycombinator.com/item?id=47741965)) "27B is a cracking, little model. It is the first small model that I, as someone with 30 years of experience, can finally say is good enough" ([source](https://news.ycombinator.com/item?id=49953771)) - Runs well on local hardware (7 comments): "Very impressive given the cost, especially since running locally you just have one time acquisition of hardware and then electricity." ([source](https://news.ycombinator.com/item?id=50228323)) "Not very fast at 20 tps (8-bit quant for total memory usage around 30 GB), but enough to feel that I have a true local coding buddy." ([source](https://news.ycombinator.com/item?id=49310121)) "The model (Qwen 3.8 27b) that I run on my own hardware on a single, high end gaming GPU, is often compared to the state of the art" ([source](https://news.ycombinator.com/item?id=49466068)) - Handles well-specified implementation tasks (3 comments): "If you just need to implement Python code from an existing plan Qwen is your choice but it has problem succeeding with more complex tasks" ([source](https://news.ycombinator.com/item?id=48647337)) "it's not great at diagnosing issues in the code, but if I can clearly articulate a test suite or boilerplate refactoring it works fine" ([source](https://news.ycombinator.com/item?id=48624885)) "This is why I have switched nearly all of my personal coding experiments over to Qwen3.6 27B." ([source](https://news.ycombinator.com/item?id=48247535)) - Fast and cheap to run (4 comments): "It feels very snappy and unlike my experience with running 27B models the performance stays pretty flat even as the context increases" ([source](https://news.ycombinator.com/item?id=49757051)) "Very impressive given the cost, especially since running locally you just have one time acquisition of hardware and then electricity." ([source](https://news.ycombinator.com/item?id=50228323)) "tiny, highly cached, fast models like Qwen's 27b/a3b range or DeepSeek flash are highly capable and relatively inexpensive to run" ([source](https://news.ycombinator.com/item?id=49426636)) ## What people warn about - Overthinks and burns tokens (4 comments): "It's definitely burning through a lot of tokens to ruminate about aspects of tasks that earlier models get done better seemingly on memory." ([source](https://news.ycombinator.com/item?id=49305977)) "With thinking on it just ruminates forever and with thinking off it gives obviously bad borderline hallucinating advice." ([source](https://news.ycombinator.com/item?id=49409833)) "It doubts itself so much and questions its every decision to the point that it'll literally blow the entire context on thinking alone" ([source](https://news.ycombinator.com/item?id=49422292)) - Weak on complex, subtle tasks (4 comments): "the smaller parameter models are not really comparable to Opus when it comes to making intelligent decisions about greyer areas of good software architecture" ([source](https://news.ycombinator.com/item?id=49452733)) "If you just need to implement Python code from an existing plan Qwen is your choice but it has problem succeeding with more complex tasks" ([source](https://news.ycombinator.com/item?id=48647337)) "it's not great at diagnosing issues in the code, but if I can clearly articulate a test suite or boilerplate refactoring it works fine" ([source](https://news.ycombinator.com/item?id=48624885)) - Censorship on sensitive topics (3 comments): "until I can ask their models what happened during the student protests in 1989 I'm just not able to trust these models with anything of substance" ([source](https://news.ycombinator.com/item?id=48210616)) "The recent Qwen 3.5 9B model is quite good, though it did trip all over itself to avoid telling me what happened on Tiananmen Square in 1989." ([source](https://news.ycombinator.com/item?id=47583593)) "This looks like it's coming from a separate "safety mechanism". Remains to be seen how much censorship is baked into the weights." ([source](https://news.ycombinator.com/item?id=46769177)) - Falls short of frontier models (3 comments): "It's a good model, but it hallucinates way more than the frontier models and often diverges even from the very explicit instructions." ([source](https://news.ycombinator.com/item?id=50083471)) "How convenient of them to compare themselves to the last generation Opus and GPT models to make their model look better than it really is." ([source](https://news.ycombinator.com/item?id=47615314)) "Nowhere near the power of ChatGPT 5.4 Pro imho... thought for maybe 15 seconds on a problem that pro would have spend 15 minutes on" ([source](https://news.ycombinator.com/item?id=47845311)) ## Alternatives hackers recommend - [Z.ai GLM](https://hackerapproved.com/p/glm-5-2.md): 1043 recommend - [Kimi](https://hackerapproved.com/p/kimi-k3.md): 931 recommend - [Gemma 4](https://hackerapproved.com/p/gemma-4.md): 713 recommend - [Qwen](https://hackerapproved.com/p/qwen.md): 474 recommend - [Qwen3.6-35B-A3B](https://hackerapproved.com/p/qwen3-6-35b-a3b.md): 373 recommend ## Cite as Hacker Approved, "Qwen 3.8 27B", https://hackerapproved.com/p/qwen-3-8-27b/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).