llama.cpp: reenable + Apriel-1.5-15b-Thinker

2025-10-16 18:44:05 -04:00
parent 3811193739
commit d96035120f
2 changed files with 4 additions and 4 deletions
--- a/services/llama-cpp.nix
+++ b/services/llama-cpp.nix
@@ -11,8 +11,8 @@
    enable = true;
    model = builtins.toString (
      pkgs.fetchurl {
-        url = "https://huggingface.co/ggml-org/gpt-oss-20b-GGUF/resolve/main/gpt-oss-20b-mxfp4.gguf";
-        sha256 = "be37a636aca0fc1aae0d32325f82f6b4d21495f06823b5fbc1898ae0303e9935";
+        url = "https://huggingface.co/unsloth/Apriel-1.5-15b-Thinker-GGUF/resolve/main/Apriel-1.5-15b-Thinker-Q4_0.gguf";
+        sha256 = "4d9439b76b6f4380ab5205617c1ef3d10b0e8897146a0a7ccb7155bca1771df7";
      }
    );
    port = service_configs.ports.llama_cpp;
@@ -27,7 +27,7 @@
    );
    extraFlags = [
      "-ngl"
-      "5"
+      "8"
      "-c"
      "16384"
    ];