From 2674eccaaede3a71c53b65e29213e31655f8ae6a Mon Sep 17 00:00:00 2001
From: Adrian Gunnar Lauterer <adrian@lauterer.it>
Date: Thu, 4 Jun 2026 20:23:28 +0200
Subject: [PATCH] add llama models

---
 modules/llama-swap.nix | 28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

diff --git a/modules/llama-swap.nix b/modules/llama-swap.nix
index ecc2a5c..a0fe351 100644
--- a/modules/llama-swap.nix
+++ b/modules/llama-swap.nix
@@ -58,6 +58,7 @@
               "qwen3.6"
               "qwen3.5"
               "qwen3.5-35b-a3b"
+              "4o"
             ];
             ttl = 1800;
             macros = {
@@ -74,20 +75,22 @@
             };
           };
 
-          "gemma4" = {
+          "gemma4-26b-a4b" = {
             cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -fa 0 -ngl $\{ngl\} --hf-repo $\{hf_repo\}   $\{tools\}";
-            ttl = 900;
+            aliases = [
+              "gemma4"
+            ];
+	    ttl = 900;
             macros = {
               hf_repo = "unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_XXS";
               ctx = 36000;
             };
           };
-
-          "LFM2.5-8B-A1B" = {
-            cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\}  $\{kv_cache\} -ngl $\{ngl\} --hf-repo $\{hf_repo\} $\{tools\} ";
-            ttl = 900;
+          "gemma4-12b" = {
+            cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -fa 0 -ngl $\{ngl\} --hf-repo $\{hf_repo\}   $\{tools\}";
+	    ttl = 900;
             macros = {
-              hf_repo = "unsloth/LFM2.5-8B-A1B-GGUF";
+              hf_repo = "unsloth/gemma-4-12B-it-GGUF";
               ctx = 128000;
             };
           };
@@ -101,12 +104,15 @@
             };
           };
 
-          "ministal-3-8b-reasonning" = {
-            cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} --hf-repo $\{hf_repo\} $\{tools\} ";
-            aliases = [ "ministral3" ];
+          "LFM2.5-8B-A1B" = {
+            cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\}  $\{kv_cache\} -ngl $\{ngl\} --hf-repo $\{hf_repo\} $\{tools\} ";
             ttl = 900;
-            macros.hf_repo = "mistralai/Ministral-3-8B-Reasoning-2512-GGUF";
+            macros = {
+              hf_repo = "LiquidAI/LFM2.5-8B-A1B-GGUF";
+              ctx = 128000;
+            };
           };
+
           "ministal-3-3b" = {
             cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -ngl $\{ngl\} --hf-repo $\{hf_repo\}  $\{tools\} ";
             aliases = [ "ministral3-mini" ];