x

config.json

@@ -20,7 +20,7 @@
   "revision": "",
   "models": {
     "default": "Qwen3.5-35B-A3B-GPTQ-Int4",
-    "selected": "Qwen3.6-35B-A3B-FP8",
+    "selected": "Qwen3.6-27B",
     "profiles": {
       "Qwen3.6-35B-A3B-FP8": {
         "local_path": "Qwen3.6-35B-A3B-FP8",
@@ -43,8 +43,9 @@
       },
       "Qwen3.6-27B": {
         "local_path": "Qwen3.6-27B",
-        "dtype": "auto",
-        "ctx": "131072",
+        "ctx": "262144",
+        "max_tokens": "32768",
+        "max_num_batched_tokens": 16384,
         "trust_remote": true,
         "enforce_eager": false,
         "valid_tp": [
@@ -52,11 +53,9 @@
           2
         ],
         "max_num_seqs": 32,
-        "max_tokens": "16384",
         "gpu_util": "0.90",
         "kv_cache_dtype": "fp8",
         "enable_prefix_caching": true,
-        "max_num_batched_tokens": 16384,
         "tool_call_parser": "qwen3_coder",
         "reasoning_parser": "qwen3",
         "enable_auto_tool_choice": true,