> Some experimental support for using Gemma 3's vision was introduced here: [#12344](https://github.com/ggml-org/llama.cpp/pull/12344) > > Tested it just now and it works flawlessly, wish it was supported on the server... _Originally posted by @ikcikoR in [#8010](https://github.com/ggml-org/llama.cpp/issues/8010#issuecomment-2721467932)_