以下部分引用 AI 总结，现阶段 AI 仍然有幻觉。请以内容的原文为准。

数据新鲜度群聊 07-14 ✓卖方 07-14 ✓Wrap 断档15天 ✕News 07-15 ✓X 断档12天 ✕页面生成 07-15

Qwen 3.6

9 atoms · 跨 6 天 · 首见 2026-04-16 · 最近 2026-06-13

三色: 🟦 fact 5 · 🟥 take 4 · stance ▲1/▼1/◆3

来源: X 9

时态: fresh:6 · aging:3

标签: 好数字:5 · 好观点:4 · 好思考:1

2026-04-19 [老化中] · scaling with hardware
a 35B model won't become better if you run it on a B200 or even B300 lol
tags: 好观点·好思考 · → daily
2026-06-13 · 27B 模型能力强于
qwen 3.6 27b is better than opus 4.5 and that runs locally on a 3090 right now.
tags: 好观点 · → daily
value: qty=better than · date=2026-06-13
2026-05-11 [老化中] · 输出质量
This is about as good as the output is (completely non playable) from qwen 3.6
tags: 好观点 · → daily
value: direction=negative
2026-04-19 [老化中] · quantization resistance
3.6 incidentally seems very resistant to quantization so you wouldn't have lost much
tags: 好观点 · → daily

2026-05-01 · coding task performance: tokens per second
qwen 3.6 27b: 32 tok/s, 18m 04s, 33,946 tokens, better visuals
tags: 好数字 · → daily
value: qty=32 tok/s
2026-05-01 · coding task total completion time
qwen 3.6 27b: 32 tok/s, 18m 04s, 33,946 tokens, better visuals
tags: 好数字 · → daily
value: qty=18m 04s
2026-05-01 · coding task output tokens used
qwen 3.6 27b: 32 tok/s, 18m 04s, 33,946 tokens, better visuals
tags: 好数字 · → daily
value: qty=33,946 tokens
2026-04-25 · model parameters
Qwen 3.6 27B (Dense, 27B Active Param/Tok)
tags: 好数字 · → daily
value: qty=27B dense
2026-04-16 · 模型参数规模
A 35B MoE vision-language model with 3B active parameters
tags: 好数字 · → daily
value: qty=35B total, 3B active · date=2026-04-16