Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
clear_view
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
by
clear_view
2y ago
deepseek-r1:14b/mistral-small:24b/qwen2.5-coder:14b fit 16GB VRAM with fast generation. 32b versions bleed into RAM and take a serious performance hit but still usable.
2.
▲
by
clear_view
2y ago
32B don't fully fit 16GB of VRAM. Still fine for higher quality answers, worth the extra wait in some cases.