Join Nostr
2026-10-11 22:47:19 UTC
in reply to

Daniel Wigton on Nostr: Wish I could do that. But I only have the one graphics card. Ollama does a decent job ...

Wish I could do that. But I only have the one graphics card. Ollama does a decent job at hotswapping models but I can go much faster with llama.cpp trouble is llama.cpp doesn't like sharing vram. It is all or crash. I'd have to actually start and stop the process with different models. Might as well just use qwen3.8 27b for everything.

In practice I am using cursor $20/mo for real development.