--n-cpu-ffn sounds interesting for dense models, I might have to try that maybe with Gemma 4 12b in my pitiful 8GB VRAM lol

]]>https://forum.ieu.app/topic/d7d04e0a-7752-4e30-af91-2f5b8050aabb/release-v0.4.0-ggml-org-llama.cppRSS for NodeSun, 06 Sep 2026 04:29:06 GMTFri, 04 Sep 2026 20:12:15 GMT60