I benchmarked 21 Qwen3.8 27B variants on 16GB VRAM

NVIDIA模型发布开源代码端侧推理

qwen3 16gb

时间与来源

收录: 09/04 21:00
来源类型: 未分类

讨论趋势

→ 平稳

最近 24 小时与此前 24 小时的快照均值对比 · 7 天曲线

百分比基于采集到的讨论信号，不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。

AI 摘要

A user benchmarked 21 Qwen3.8 27B variants on a 16GB VRAM GPU (RTX 5080) using C code. The results showed varying performance among the models, with some quants being underwhelming. The benchmark data includes Model Mean KLD, Same top p, and GGUF size for each variant. Notably, two unsloth/Qwen3.8-27B-UD-Q4_K_XL variants (UD3 and UD2) could not fit on the 16GB VRAM, having sizes of 16.4GiB and 16.7GiB respectively.