跳到正文
RCreddit.com·
暂不在当前实时榜单

Qwen3.5 0.8B on CPU

AI 摘要

The Qwen3.5 0.8B model's performance on CPUs was evaluated for local dictation cleanup. A specialized engine, qwen35-cpu, demonstrated significant improvements over llama.cpp running Unsloth Q4_0. Specifically, qwen35-cpu achieved approximately 2.9x prefill, 1.3x single-request decode, and 1.7x batch-16 throughput. This makes the Qwen3.5 0.8B model a viable option for CPU-based tasks, especially when GPUs are occupied with other processes.

时间与来源

时间显示为 UTC

显示时区:UTC

本地时区尚不可用,暂时显示 UTC。

收录当时偏移:UTC+02026年9月8日 08:00 UTC

收录
2026年9月8日 08:00
来源类型
开发者社区
正文

本站未收录正文。

前往源站阅读 →
来源·reddit.com