Add SGLang serving instructions

README.md

@@ -76,6 +76,7 @@ Intern-S2-Mobius can be deployed using any of the following LLM inference framew
 - LMDeploy
 - Transformers
 - vLLM
+- SGLang
 
 ### Sampling Parameters
 
@@ -192,5 +193,32 @@ vllm serve \
     --tool-call-parser qwen3_coder 
 ```
 
+### SGLang
 
+Use the `lmsysorg/sglang:dev` Docker image or a recent source build with Intern-S2-Mobius support. See the [SGLang cookbook](https://docs.sglang.io/cookbook/autoregressive/InternLM/Intern-S2-Mobius) for Docker commands, verified deployment recipes, benchmarks, and usage examples.
 
+- Serving With MTP (Recommended)
+
+```bash
+sglang serve \
+    --model-path internlm/Intern-S2-Mobius \
+    --trust-remote-code \
+    --tp 2 \
+    --reasoning-parser qwen3 \
+    --tool-call-parser qwen3_coder \
+    --speculative-algorithm NEXTN \
+    --speculative-num-steps 3 \
+    --speculative-eagle-topk 1 \
+    --speculative-num-draft-tokens 4
+```
+
+- Basic Serving Without MTP
+
+```bash
+sglang serve \
+    --model-path internlm/Intern-S2-Mobius \
+    --trust-remote-code \
+    --tp 2 \
+    --reasoning-parser qwen3 \
+    --tool-call-parser qwen3_coder
+```