From 2d47c441fefd563a70312034d55bd7de321cdc0a Mon Sep 17 00:00:00 2001
From: Simon Gardling <titaniumtown@proton.me>
Date: Mon, 31 Mar 2025 18:33:24 -0400
Subject: [PATCH] llm: use Q4_0 quants (faster)

---
 services/llama-cpp.nix | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/services/llama-cpp.nix b/services/llama-cpp.nix
index 008c904..59f67e3 100644
--- a/services/llama-cpp.nix
+++ b/services/llama-cpp.nix
@@ -26,8 +26,8 @@ in
     enable = true;
     model = builtins.toString (
       pkgs.fetchurl {
-        url = "https://huggingface.co/bartowski/google_gemma-3-12b-it-GGUF/resolve/main/google_gemma-3-12b-it-IQ4_XS.gguf";
-        sha256 = "aa7b7ae0b17931c379ede82da59b01f246046925aeb752af1ab4285a3b0d69db";
+        url = "https://huggingface.co/bartowski/google_gemma-3-12b-it-GGUF/resolve/main/google_gemma-3-12b-it-Q4_0.gguf";
+        sha256 = "9a7b70be8727da9fb28523b35946dd42d4fe0f622cce03daa44fccff0775516d";
       }
     );
     port = service_configs.ports.llama_cpp;