Qwen 3.8 27B at ~3 BPW on an RTX 3060: GSQ vs ByteShape IQ3-XXS 2.88BPW
英文标题：Qwen 3.8 27B at ~3 BPW on an RTX 3060: GSQ vs ByteShape IQ3-XXS 2.88BPW
Someone recommended that I try the ByteShape Qwen 3.8 27B IQ3-XXS GGUF after seeing my previous testing of the GSQ quant. So I did. And the result was… surprisingly bad. For context, I'm running: RTX 3060 12GB 16GB DDR4 RAM, single channel CachyOS / Arch Linux llama.cpp Qwen 3.8 27B MTP/specu
打开原文相关 2 · 机会 0
中文
Someone recommended that I try the ByteShape Qwen 3.8 27B IQ3-XXS GGUF after seeing my previous testing of the GSQ quant. So I did. And the result was… surprisingly bad. For context, I'm running: RTX 3060 12GB 16GB DDR4 RAM, single channel CachyOS / Arch Linux llama.cpp Qwen 3.8 27B MTP/specu
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
Someone recommended that I try the ByteShape Qwen 3.8 27B IQ3-XXS GGUF after seeing my previous testing of the GSQ quant. So I did. And the result was… surprisingly bad. For context, I'm running: RTX 3060 12GB 16GB DDR4 RAM, single channel CachyOS / Arch Linux llama.cpp Qwen 3.8 27B MTP/specu