[docs] fix sglang serve --model-path in cookbooks (#23905)

docs_new/cookbook/autoregressive/DeepSeek/DeepSeek-V3_2.mdx

@@ -55,7 +55,7 @@ DeepSeek-V3.2 supports reasoning mode. Enable the reasoning parser during deploy
 
 ```shell Command
 sglang serve \
-  --model deepseek-ai/DeepSeek-V3.2-Exp \
+  --model-path deepseek-ai/DeepSeek-V3.2-Exp \
   --reasoning-parser deepseek-v3 \
   --tp 8 \
   --host 0.0.0.0 \
@@ -139,7 +139,7 @@ For DeepSeek-V3.2-Exp:
 
 ```shell Command
 sglang serve \
-  --model deepseek-ai/DeepSeek-V3.2-Exp \
+  --model-path deepseek-ai/DeepSeek-V3.2-Exp \
   --tool-call-parser deepseekv31 \
   --reasoning-parser deepseek-v3 \
   --chat-template ./examples/chat_template/tool_chat_template_deepseekv32.jinja \

docs_new/cookbook/autoregressive/GLM/GLM-5.mdx

@@ -101,7 +101,7 @@ Deploy GLM-5 with the following command (FP8 on H200, all features enabled):
 
 ```shell Command
 sglang serve \
-  --model zai-org/GLM-5-FP8 \
+  --model-path zai-org/GLM-5-FP8 \
   --tp 8 \
   --tool-call-parser glm47 \
   --reasoning-parser glm45 \
@@ -120,7 +120,7 @@ The following ROCm command is an additional option for AMD GPUs and does not rep
 
 ```shell Command
 sglang serve \
-  --model zai-org/GLM-5 \
+  --model-path zai-org/GLM-5 \
   --tp 8 \
   --trust-remote-code \
   --nsa-prefill-backend tilelang \

docs_new/src/snippets/autoregressive/deepseek-v32-deployment.jsx

@@ -174,7 +174,7 @@ export const DeepSeekV32Deployment = () => {
     // NVFP4: fixed config
     if (isNvfp4) {
       let cmd = 'sglang serve \\\n';
-      cmd += `  --model ${modelName}`;
+      cmd += `  --model-path ${modelName}`;
       cmd += ' \\\n  --tp 4';
       cmd += ' \\\n  --quantization modelopt_fp4';
       cmd += ' \\\n  --moe-runner-backend flashinfer_trtllm';
@@ -184,14 +184,14 @@ export const DeepSeekV32Deployment = () => {
     // MXFP4: fixed config for AMD
     if (isMxfp4) {
       let cmd = 'sglang serve \\\n';
-      cmd += `  --model ${modelName}`;
+      cmd += `  --model-path ${modelName}`;
       cmd += ' \\\n  --tp 8';
       cmd += ' \\\n  --trust-remote-code';
       return cmd;
     }
 
     let cmd = 'sglang serve \\\n';
-    cmd += `  --model ${modelName}`;
+    cmd += `  --model-path ${modelName}`;
 
     // Hardware platform specific parameters
     if (isAMD) {

docs_new/src/snippets/autoregressive/glm-5-deployment.jsx

@@ -152,7 +152,7 @@ export const GLM5Deployment = () => {
     const memFraction = hwConfig.mem;
 
     let cmd = 'sglang serve \\\n';
-    cmd += `  --model ${modelName}`;
+    cmd += `  --model-path ${modelName}`;
     cmd += ` \\\n  --tp ${tpValue}`;
 
     // NVFP4 B200: trtllm NSA backends, flashinfer fusion, FP8 KV cache.

docs_new/src/snippets/autoregressive/ministral-3-deployment.jsx

@@ -50,7 +50,7 @@ export const Ministral3Deployment = () => {
 
     let cmd = 'sglang serve \\\n';
 
-    cmd += `  --model ${modelCfg.modelId}`;
+    cmd += `  --model-path ${modelCfg.modelId}`;
 
     if (tp > 1) {
       cmd += ` \\\n  --tp ${tp}`;