Qwen 3.6: what developers actually say
We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on Qwen 3.6. 79 of the 94 opinions come from people who say they use it themselves.
Developers give Qwen 3.6 76% approval across 94 opinions from 80 authors, with 79 firsthand reports. Approval over the last 90 days is 71%, below the overall figure. Monthly samples are uneven: 82% in April, 72% in June on 37 opinions, and 88% in September; the two October opinions scored 25%, too small a sample to read as a trend. Code quality (25 positive, 6 negative), local privacy (21 positive, 2 negative), and speed (13 positive, 1 negative) drive the positive case; reliability is split at 4 and 4.
Pick it if
- Want local inference for privacy-sensitive work
- Value fast answers on a Mac
- Need a low-cost local model for simple tasks
Look elsewhere if
- Need reliable agentic coding on large repos
- Plan to run quantized local builds for agent tasks
- Need a local model for moderately complex coding
Strengths and weaknesses
Each row counts opinions that mentioned the topic. Bars to the right of the line are praise, bars to the left are complaints.
How opinion has moved
What developers say
Praise
Complaints
Summaries are paraphrased. Each links to the original comment.
Qwen 3.6 compared
FAQ
Is Qwen 3.6 worth it?
Developers give it 76% approval, and 79 of 94 opinions are firsthand. It suits local, cost-sensitive work with strong privacy and speed scores. Reliability is even at 4 positive and 4 negative, so test it on your own workload.
Is Qwen 3.6 better than Gemma 4?
The preference data is mixed. Three developers preferred Qwen 3.6 over Gemma 4, while six preferred Gemma 4. Qwen 3.6 has strong local privacy and speed scores, so the better choice depends on your task.
Is Qwen 3.6 good for coding?
Code quality is its strongest aspect at 25 positive and 6 negative opinions. Developers report correct, concise answers and tested code generation. Others cite misdiagnosed bugs, slow code, context exhaustion on a 30M+ line repo, and weak quantized local results.