Update rtx6000-vllm.yaml

infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml

@@ -39,7 +39,7 @@ spec:
             - --tokenizer-mode=hf 
             - --gpu-memory-utilization=0.90  
             #- --enforce-eager 
-            - --max-model-len=98304 
+            - --max-model-len=136876
             - --enable-auto-tool-choice 
             - --kv-cache-dtype=fp8 
             - --max-num-batched-tokens=32768 
@@ -51,6 +51,7 @@ spec:
             - --enable-prefix-caching
             - --tool-call-parser=qwen3_xml
             - --reasoning-parser=qwen3
+            - --speculative-config '{"method": "mtp", "num_speculative_tokens": 2}'
             
           ports:
             - containerPort: 8000