Qwen
54.0
比平时降智
体感分
▲ 0.7
置信度: Medium
本周共追踪 813 条意见 · 7 日滚动窗口
30 天体感分曲线
版本细分
版本分只统计明确点名该版本的评论。
| Qwen3.6 27B | 53.2−0.9 | n=288 | 置信度: Medium | |
|---|---|---|---|---|
| Qwen3.6 35B A3B | 65.6+11.6 | n=45 | 置信度: Medium | |
| Qwen3.5-122B-A10B | 72.2+18.1 | n=27 | 置信度: Medium | |
| Qwen3.5-9B | 64.1+10.0 | n=8 | 置信度: Low | |
| Qwen3 Coder Next | 61.4+7.4 | n=6 | 置信度: Low |
另有 424 条泛称评论计入家族分，不计入任何版本分。
分维度体感
| 通用文本 | 57 | n=228 | |
|---|---|---|---|
| 编程 | 58 | n=163 | |
| 推理 | 60 | n=154 | |
| 图像/视觉 | 66 | n=11 | |
| 角色扮演/创作 | 样本不足 | n=3 | |
| 安全与拒答 | 样本不足 | n=1 | |
| 速度与延迟 | 65 | n=217 | |
| 本地部署 | 66 | n=36 |
主要好评因素
- speed quality 速度与延迟 ×109
- strong reasoning 推理 ×63
- problem solved 通用文本 ×27
- competitor flight 通用文本 ×20
- problem solved 编程 ×19
主要差评因素
- performance issue 速度与延迟 ×25
- speed quality 速度与延迟 ×21
- competitor flight 通用文本 ×15
- tool failure 通用文本 ×15
- performance issue 编程 ×14
用户原声
🟢 “I'm running it with preserve thinking now, and it is not doing stupid things, so this appears to be a huge improvement”
来自 reddit · text_general · en
🟢 “It's truly impressive what some people have done with comparatively tiny models like Qwen3.6-27B, just by using clever prompting to have it break up tasks and make it not rely on its built-in knowledge as much.”
来自 reddit · reasoning · en
🔴 “I was shocked at how stupid qwen 3.5 122b was at 4 bit compared to qwen 3.6 27b at 8 bit. Night and day.”
来自 reddit · text_general · en
🔴 “Yeah, without preserve thinking it forgets what it concluded earlier and usually goes down the same paths again. It's a bit odd it's not on by default.”
来自 reddit · text_general · en
🟢 “accurate enough for the speed”
来自 reddit · speed_latency · en