x

config.json

@@ -20,7 +20,7 @@
   "revision": "",
   "models": {
     "default": "Qwen3.5-35B-A3B-GPTQ-Int4",
-    "selected": "Qwen3.5-27B-FP8",
+    "selected": "Gemma-4-26B-A4B",
     "profiles": {
       "GLM-4.7-Flash-AWQ": {
         "local_path": "GLM-4.7-Flash-AWQ",
@@ -95,7 +95,7 @@
         "ctx": "131072",
         "trust_remote": true,
         "valid_tp": [2],
-        "max_num_seqs": "6",
+        "max_num_seqs": "12",
         "max_tokens": "8096",
         "gpu_util": "0.90",
         "enforce_eager": false,
@@ -104,6 +104,22 @@
         "enable_auto_tool_choice": true,
         "served_model_name": "Qwen3-Next-80B-A3B-Instruct-AWQ-4bit",
         "hf_model_id": "cpatonn/Qwen3-Next-80B-A3B-Instruct-AWQ-4bit"
+      },
+      "Gemma-4-26B-A4B": {
+        "local_path": "gemma-4-26B-A4B",
+        "dtype": "auto",
+        "ctx": "65536",
+        "trust_remote": true,
+        "valid_tp": [1, 2],
+        "max_num_seqs": 16,
+        "max_tokens": "8192",
+        "enforce_eager": false,
+        "gpu_util": "0.85",
+        "kv_cache_dtype": "fp8",
+        "enable_prefix_caching": true,
+        "max_num_batched_tokens": 16384,
+        "served_model_name": "Gemma-4-26B-A4B",
+        "hf_model_id": "google/gemma-4-26B-A4B"
       }
     }
   }