Qwen3.8-27B在双RTX 3060上跑出40 tok/s，实测CUDA编程能力
anderspitman · reddit · 2026-08-15
Reddit user anderspitman shares early performance report of Qwen3.8-27B on 2x RTX 3060 12GB. With specific llama.cpp config, 40 tok/s achieved. In a CUDA vector add smoke test, the model one-shot generated the program and installed compiler, but couldn't run due to VRAM limits. It attempted debugging, and eventually succeeded after shutting down llama-server.
所属事件:双 RTX 3060 实测 Qwen3.8-27B 跑出 50 tok/s(2 条相关)→
「编程与Agent」频道最新
- 免 Docker 的内核级 agent 沙盒:macOS Seatbelt 实践 — mostly_deterministic · 2026-10-16
- 把 Agent 工作从 GPU 挪回 CPU，让动作更持久可预测 — cramforce · 2026-10-16
- Jev 教程：用 180ms 分类模型替代「PASS/FAIL」字符串匹配做工作流判断 — max_paperclips · 2026-10-16
- 别给 Agent 裸 API Key：预付余额加凭证级限额才是正解 — Cold_Hall_5384 · 2026-10-16
- 六个月预测:通用 agent 框架岗位需求将降温 — yi111 · 2026-10-16
- Omarchy 装不上 NVIDIA 机器？他直接让 Claude 修好了 ISO — indragie · 2026-10-16