diff --git a/qwen/qwen-110B-chat/config.yaml b/qwen/qwen-110B-chat/config.yaml
new file mode 100644
index 00000000..ed7daef4
--- /dev/null
+++ b/qwen/qwen-110B-chat/config.yaml
@@ -0,0 +1,16 @@
+external_package_dirs: []
+model_metadata:
+  example_model_input: {"prompt": "How long would it take to reach the sun?"}
+model_name: Qwen1.5-vllm-streaming
+python_version: py310
+requirements:
+- torch==2.2.1
+- transformers==4.40.0
+- vllm==0.4.1
+- asyncio==3.4.3
+- ray
+resources:
+  accelerator: A100:4
+  use_gpu: true
+secrets: {}
+system_packages: []
diff --git a/qwen/qwen-110B-chat/model/__init__.py b/qwen/qwen-110B-chat/model/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/qwen/qwen-110B-chat/model/model.py b/qwen/qwen-110B-chat/model/model.py
new file mode 100644
index 00000000..a58fcd07
--- /dev/null
+++ b/qwen/qwen-110B-chat/model/model.py
@@ -0,0 +1,65 @@
+import subprocess
+import uuid
+from transformers import AutoTokenizer
+
+from vllm import SamplingParams
+from vllm.engine.arg_utils import AsyncEngineArgs
+from vllm.engine.async_llm_engine import AsyncLLMEngine
+
+
+class Model:
+    def __init__(self, model_name="Qwen/Qwen1.5-110B-Chat"):
+        self.model_name = model_name
+        self.tokenizer = None
+        self.sampling_params = None
+
+        command = "ray start --head"
+        subprocess.check_output(command, shell=True, text=True)
+
+    def load(self):
+        self.model_args = AsyncEngineArgs(
+            model=self.model_name,
+            dtype='auto',
+            enforce_eager=True,
+            tensor_parallel_size=4
+
+        )
+
+        self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)
+
+        self.sampling_params = SamplingParams(    # Using default values
+            temperature=0.7,
+            top_p=0.8,
+            repetition_penalty=1.05,
+            max_tokens=512
+        )
+
+        self.llm_engine = AsyncLLMEngine.from_engine_args(self.model_args)
+
+    async def predict(self, model_input):
+        message = model_input.pop("prompt")
+
+        prompt = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": message}
+        ]
+
+        text = self.tokenizer.apply_chat_template(
+            prompt,
+            tokenize=False,
+            add_generation_prompt=True
+        )
+
+        idx = str(uuid.uuid4().hex)
+        vllm_generator = self.llm_engine.generate(text, self.sampling_params, idx)
+
+        async def generator():
+            full_text = ""
+            async for output in vllm_generator:
+                text = output.outputs[0].text
+                delta = text[len(full_text) :]
+                full_text = text
+                yield delta
+
+        return generator()
+