← 全部模型 Qwen · 近 7 天Qwen3.6 35B A3B0–100 分，越高代表社区使用体验越正面，不代表能力测试成绩。更新于 2026-09-05 17:31 UTC · 2026-08-29 – 2026-09-05 UTC社区口碑分64.2/10033 条评论 · 近 7 天 社区评论精选近 7 天的不同意见，摘录条数不代表真实好评比例。全部摘录正面负面1 条精选摘录 中性速度与延迟2026-09-05 “qwen 3.6 35b a3b 在 80+ tok/s” Redditen机器翻译 展开原文qwen 3.6 35b a3b at 80+ tok/s负面编程2026-09-05 “Qwen3.6 35B和其他微调版本与Minimax M2.7有类似行为。相当于略差的M2.7。” Redditen机器翻译 展开原文Qwen3.6 35B and other finetunes have similar behaviour to Minimax M2.7. It's like a slightly worse M2.7.中性编程2026-09-05 “它做到了。但所有图形都是方块和圆形。” Redditen机器翻译 展开原文It made it. But all the graphics were squares and circles.正面速度与延迟2026-09-04 “那台机器的话，我会从Qwen3.6 35b a3b的Q4量化版开始。 MLX不再默认最强，但试试oMLX和llama.cpp（通过LM Studio等），看看效果如何。” Redditen机器翻译 展开原文Q4-ish quant of Qwen3.6 35b a3b would be my starting point with that machine. MLX is no longer king by default, however, try oMLX and llama.cpp (via LM Studio etc) and see what happens.负面编程2026-09-04 “但我的Q4_K_S Qwen3.6 35B A3B连最简单的C++ HTTP服务器都写不出来。我不明白我哪里做错了” Redditen机器翻译 展开原文Yet my Q4_K_S Qwen3.6 35B A3B cant even write the simplest C++ HTTP server. I dont understand what im doing wrong正面编程2026-09-03 “你提到的其他3个模型都不可能写出接近OP展示的那个游戏的代码” Redditen机器翻译 展开原文the other 3 models you mentioned aren't going to be capable of coding anything even close to the game OP was showing负面通用文本2026-09-03 “模型本身能力没那么强，量化会降低准确性/增加幻觉，而KV缓存量化会进一步降低精度。” Redditen机器翻译 展开原文The model itself isn't that capable, the quantization reduces accuracy/increases hallucinations, and KV cache quantization further reduces precision.中性本地部署2026-09-03 “我有一台48GB的M4 Pro，但用它跑Qwen3.6 35b a3b或27b配合大上下文还是不太够。” Redditen机器翻译 展开原文I had a M4 Pro with 48GB, and it wasn't quite enough for e.g. Qwen3.6 35b a3b or 27b with large context.正面推理2026-09-03 “当把这个模型接到后台代理工作负载时，它远不如训练良好的Qwen 3.6 35B-A3B” Redditen机器翻译 展开原文when I hook this model to the background agent workload, it was far worse than the well trained Qwen 3.6 35B-A3B正面编程2026-09-02 “35B模型会进行审查、运行测试，并优化以获得更好的性能、可读性和更小的占用空间” Redditen机器翻译 展开原文the 35B model reviews it, runs tests, and optimizes it for better performance, readability, and a smaller footprint正面推理2026-09-01 “每一款都是升级” Redditen机器翻译 展开原文Each is a step up正面通用文本2026-09-01 “Qwen 3.6 35b确实能完成任务，但天哪它确实显得干巴巴的” Redditen机器翻译 展开原文Qwen 3.6 35b actually gets the job done but boy does it come off dry中性通用文本2026-09-01 “标准版Qwen3.6 35b” Redditen机器翻译 展开原文standard Qwen3.6 35b负面通用文本2026-08-31 “你得跟他们说好几遍才能修一个问题” Redditen机器翻译 展开原文you had to tell them several times just to fix one problem负面通用文本2026-09-03 “他们的36b-a4b在除了一项之外的所有基准测试中都击败了qwen3.6-35b-a3b” Redditen机器翻译 展开原文Their 36b-a4b beats qwen3.6-35b-a3b on all but one of those benchmarks 近 7 天暂时没有符合此筛选条件的精选评论。查看更多摘录
负面编程2026-09-05 “Qwen3.6 35B和其他微调版本与Minimax M2.7有类似行为。相当于略差的M2.7。” Redditen机器翻译 展开原文Qwen3.6 35B and other finetunes have similar behaviour to Minimax M2.7. It's like a slightly worse M2.7.
中性编程2026-09-05 “它做到了。但所有图形都是方块和圆形。” Redditen机器翻译 展开原文It made it. But all the graphics were squares and circles.
正面速度与延迟2026-09-04 “那台机器的话，我会从Qwen3.6 35b a3b的Q4量化版开始。 MLX不再默认最强，但试试oMLX和llama.cpp（通过LM Studio等），看看效果如何。” Redditen机器翻译 展开原文Q4-ish quant of Qwen3.6 35b a3b would be my starting point with that machine. MLX is no longer king by default, however, try oMLX and llama.cpp (via LM Studio etc) and see what happens.
负面编程2026-09-04 “但我的Q4_K_S Qwen3.6 35B A3B连最简单的C++ HTTP服务器都写不出来。我不明白我哪里做错了” Redditen机器翻译 展开原文Yet my Q4_K_S Qwen3.6 35B A3B cant even write the simplest C++ HTTP server. I dont understand what im doing wrong
正面编程2026-09-03 “你提到的其他3个模型都不可能写出接近OP展示的那个游戏的代码” Redditen机器翻译 展开原文the other 3 models you mentioned aren't going to be capable of coding anything even close to the game OP was showing
负面通用文本2026-09-03 “模型本身能力没那么强，量化会降低准确性/增加幻觉，而KV缓存量化会进一步降低精度。” Redditen机器翻译 展开原文The model itself isn't that capable, the quantization reduces accuracy/increases hallucinations, and KV cache quantization further reduces precision.
中性本地部署2026-09-03 “我有一台48GB的M4 Pro，但用它跑Qwen3.6 35b a3b或27b配合大上下文还是不太够。” Redditen机器翻译 展开原文I had a M4 Pro with 48GB, and it wasn't quite enough for e.g. Qwen3.6 35b a3b or 27b with large context.
正面推理2026-09-03 “当把这个模型接到后台代理工作负载时，它远不如训练良好的Qwen 3.6 35B-A3B” Redditen机器翻译 展开原文when I hook this model to the background agent workload, it was far worse than the well trained Qwen 3.6 35B-A3B
正面编程2026-09-02 “35B模型会进行审查、运行测试，并优化以获得更好的性能、可读性和更小的占用空间” Redditen机器翻译 展开原文the 35B model reviews it, runs tests, and optimizes it for better performance, readability, and a smaller footprint
正面通用文本2026-09-01 “Qwen 3.6 35b确实能完成任务，但天哪它确实显得干巴巴的” Redditen机器翻译 展开原文Qwen 3.6 35b actually gets the job done but boy does it come off dry
负面通用文本2026-08-31 “你得跟他们说好几遍才能修一个问题” Redditen机器翻译 展开原文you had to tell them several times just to fix one problem
负面通用文本2026-09-03 “他们的36b-a4b在除了一项之外的所有基准测试中都击败了qwen3.6-35b-a3b” Redditen机器翻译 展开原文Their 36b-a4b beats qwen3.6-35b-a3b on all but one of those benchmarks
社区评论
精选近 7 天的不同意见，摘录条数不代表真实好评比例。
1 条精选摘录
展开原文
qwen 3.6 35b a3b at 80+ tok/s
展开原文
Qwen3.6 35B and other finetunes have similar behaviour to Minimax M2.7. It's like a slightly worse M2.7.
展开原文
It made it. But all the graphics were squares and circles.
展开原文
Q4-ish quant of Qwen3.6 35b a3b would be my starting point with that machine. MLX is no longer king by default, however, try oMLX and llama.cpp (via LM Studio etc) and see what happens.
展开原文
Yet my Q4_K_S Qwen3.6 35B A3B cant even write the simplest C++ HTTP server. I dont understand what im doing wrong
展开原文
the other 3 models you mentioned aren't going to be capable of coding anything even close to the game OP was showing
展开原文
The model itself isn't that capable, the quantization reduces accuracy/increases hallucinations, and KV cache quantization further reduces precision.
展开原文
I had a M4 Pro with 48GB, and it wasn't quite enough for e.g. Qwen3.6 35b a3b or 27b with large context.
展开原文
when I hook this model to the background agent workload, it was far worse than the well trained Qwen 3.6 35B-A3B
展开原文
the 35B model reviews it, runs tests, and optimizes it for better performance, readability, and a smaller footprint
展开原文
Each is a step up
展开原文
Qwen 3.6 35b actually gets the job done but boy does it come off dry
展开原文
standard Qwen3.6 35b
展开原文
you had to tell them several times just to fix one problem
展开原文
Their 36b-a4b beats qwen3.6-35b-a3b on all but one of those benchmarks
近 7 天暂时没有符合此筛选条件的精选评论。