chore(llama): update llama-cpp and remove deprecated models

- Upgrade llama-cpp from version 8196 to 8229
- Remove GPT OSS CSEC (20B) - Thinking model config
- Remove Qwen3 Next (80B) - Instruct model config
- Increment -ncmoe parameter for Qwen3 Coder Next (80B) model

modules/home/programs/graphical/wms/hyprland/default.nix

@@ -1,8 +1,9 @@
-{ lib
-, pkgs
-, config
-, namespace
-, ...
+{
+  lib,
+  pkgs,
+  config,
+  namespace,
+  ...
 }:
 let
   inherit (lib) types mkIf;
@@ -92,8 +93,8 @@ in
           ",XF86AudioLowerVolume, exec, wpctl set-volume @DEFAULT_AUDIO_SINK@ 5%-"
           ",XF86AudioMute, exec, wpctl set-mute @DEFAULT_AUDIO_SINK@ toggle"
           ",XF86AudioMicMute, exec, wpctl set-mute @DEFAULT_AUDIO_SOURCE@ toggle"
-          ",XF86MonBrightnessUp, exec, brightnessctl s 10%+"
-          ",XF86MonBrightnessDown, exec, brightnessctl s 10%-"
+          ",XF86MonBrightnessUp, exec, brightnessctl s 4%+"
+          ",XF86MonBrightnessDown, exec, brightnessctl s 5%-"
 
           # macOS Keyboard Brightness
           "$menuMod, XF86MonBrightnessUp, exec, brightnessctl -d kbd_backlight s 10%+"

modules/nixos/services/llama-swap/config.nix

@@ -73,51 +73,6 @@ in
       };
     };
 
-    # https://huggingface.co/mradermacher/GPT-OSS-Cybersecurity-20B-Merged-i1-GGUF/tree/main
-    "gpt-oss-csec-20b-thinking" = {
-      name = "GPT OSS CSEC (20B) - Thinking";
-      macros.ctx = "131072";
-      cmd = ''
-        ${llama-cpp}/bin/llama-server \
-          --port ''${PORT} \
-          -m /mnt/ssd/Models/GPT-OSS/GPT-OSS-Cybersecurity-20B-Merged.i1-MXFP4_MOE.gguf \
-          -c ''${ctx} \
-          --temp 1.0 \
-          --top-p 1.0 \
-          --top-k 40 \
-          -dev CUDA0
-      '';
-      metadata = {
-        type = [ "text-generation" ];
-      };
-    };
-
-    # https://huggingface.co/unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF/tree/main
-    "qwen3-next-80b-instruct" = {
-      name = "Qwen3 Next (80B) - Instruct";
-      macros.ctx = "262144";
-      cmd = ''
-        ${llama-cpp}/bin/llama-server \
-          --port ''${PORT} \
-          -m /mnt/ssd/Models/Qwen3/Qwen3-Next-80B-A3B-Instruct-UD-Q2_K_XL.gguf \
-          -c ''${ctx} \
-          --temp 0.7 \
-          --min-p 0.0 \
-          --top-p 0.8 \
-          --top-k 20 \
-          --repeat-penalty 1.05 \
-          -ctk q8_0 \
-          -ctv q8_0 \
-          -fit off
-      '';
-      metadata = {
-        type = [
-          "text-generation"
-          "coding"
-        ];
-      };
-    };
-
     # https://huggingface.co/unsloth/Qwen3-Coder-Next-GGUF/tree/main
     "qwen3-coder-next-80b-instruct" = {
       name = "Qwen3 Coder Next (80B) - Instruct";
@@ -132,7 +87,7 @@ in
           --min-p 0.01 \
           --top-k 40 \
           -fit off \
-          -ncmoe 18 \
+          -ncmoe 19 \
           -ts 78,22
       '';
 

packages/llama-cpp/default.nix

@@ -7,12 +7,12 @@
   vulkanSupport = true;
 }).overrideAttrs
   (oldAttrs: rec {
-    version = "8196";
+    version = "8229";
     src = pkgs.fetchFromGitHub {
       owner = "ggml-org";
       repo = "llama.cpp";
       tag = "b${version}";
-      hash = "sha256-GZRHiyT8mvhV5RTczDRnCSh31UxRZ3F8tEBC1l8oFNQ=";
+      hash = "sha256-SmCNsQfLQMmwa8PzFPaQb9yBdUZTxM8xxSqhumVGvHM=";
       leaveDotGit = true;
       postFetch = ''
         git -C "$out" rev-parse --short HEAD > $out/COMMIT