4 ms·
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
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.