Hacker News Reader

If I understood it correctly: this time no, it is actually new engine builded by the ollama team indipendent from llama.cpp

I doubt it. Llama.cpp just added support for the same models a few weeks ago. Folks at ollama just did a git pull.

It's open source, you could have checked. Seems indeed like the new engine cuts out llama.cpp, using GGML libary directly.

seriously? who do you think develops ggml?

hint: it's llama.cpp

llama.cpp added support for vision 6 days ago.

See SimonW post here:

>If I understood it correctly

You understood it exactly like they wanted you to...