diff --git a/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/README.md b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/README.md
new file mode 100644
index 00000000..a40a3a5e
--- /dev/null
+++ b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/README.md
@@ -0,0 +1,59 @@
+# SmolLM2-135M-Instruct Optimization Recipe for CPU
+
+This folder contains the optimization recipe for the [SmolLM2-135M-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM2-135M-Instruct) model targeting CPU execution. The model is optimized to **INT4** precision using Microsoft Olive and ONNX Runtime GenAI's ModelBuilder.
+
+## 📊 Recipe Details
+
+| Property | Details |
+| :--- | :--- |
+| **Model Name** | `HuggingFaceTB/SmolLM2-135M-Instruct` |
+| **Architecture** | SmolLM2 (Llama-based) |
+| **Target Device** | CPU |
+| **Precision** | INT4 |
+| **Execution Provider** | `CPUExecutionProvider` |
+| **Optimization Tool** | Microsoft Olive (ModelBuilder) |
+
+## 🛠️ Prerequisites
+
+Before running the optimization, ensure you have the required dependencies installed.
+
+```bash
+pip install -r requirements.txt
+```
+
+## 🚀 How to Run Optimization
+
+Navigate to this directory and run the following command to optimize the model:
+
+```bash
+python -m olive run --config olive_config.json
+```
+
+This will download the model, apply INT4 quantization, and save the optimized ONNX model in the `models/smollm_manual` directory.
+
+## 🤖 How to Run Inference
+
+Once the model is optimized, you can use `onnxruntime-genai` to run inference locally.
+
+**Example Python Snippet:**
+
+```python
+import onnxruntime_genai as og
+
+model = og.Model("models/smollm_manual")
+tokenizer = og.Tokenizer(model)
+
+prompt = "<|im_start|>user\nExplain quantum physics in one sentence.<|im_end|>\n<|im_start|>assistant\n"
+tokens = tokenizer.encode(prompt)
+
+params = og.GeneratorParams(model)
+params.set_search_options(max_length=100)
+params.input_ids = tokens
+
+generator = og.Generator(model, params)
+
+while not generator.is_done():
+    generator.compute_logits()
+    generator.generate_next_token()
+    print(tokenizer.decode(generator.get_next_tokens()), end='', flush=True)
+```
\ No newline at end of file
diff --git a/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/info.yml b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/info.yml
new file mode 100644
index 00000000..d4ad42bb
--- /dev/null
+++ b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/info.yml
@@ -0,0 +1,6 @@
+arch: SmolLM2
+recipes:
+  - name: SmolLM2-135M-Instruct CPU INT4
+    file: olive_config.json
+    devices: cpu
+    eps: CPUExecutionProvider
\ No newline at end of file
diff --git a/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/olive_ci.json b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/olive_ci.json
new file mode 100644
index 00000000..2dafb7c5
--- /dev/null
+++ b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/olive_ci.json
@@ -0,0 +1,9 @@
+[
+  {
+    "name": "smollm2_cpu_int4",
+    "os": "windows",
+    "device": "cpu",
+    "requirements_file": "requirements.txt",
+    "command": "python -m olive run --config olive_config.json"
+  }
+]
\ No newline at end of file
diff --git a/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/olive_config.json b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/olive_config.json
new file mode 100644
index 00000000..e1d11161
--- /dev/null
+++ b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/olive_config.json
@@ -0,0 +1,36 @@
+{
+    "input_model": {
+      "type": "HfModel",
+      "config": {
+        "model_path": "HuggingFaceTB/SmolLM2-135M-Instruct"
+      }
+    },
+    "systems": {
+      "local_system": {
+        "type": "LocalSystem",
+        "config": {
+          "accelerators": [
+            {
+              "device": "cpu",
+              "execution_providers": ["CPUExecutionProvider"]
+            }
+          ]
+        }
+      }
+    },
+    "passes": {
+      "builder": {
+        "type": "ModelBuilder",
+        "config": {
+          "precision": "int4"
+        }
+      }
+    },
+    "engine": {
+      "log_severity_level": 1,
+      "host": "local_system",
+      "target": "local_system",
+      "cache_dir": ".olive-cache",
+      "output_dir": "models/smollm_manual"
+    }
+  }
\ No newline at end of file
diff --git a/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/requirements.txt b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/requirements.txt
new file mode 100644
index 00000000..50c07af4
--- /dev/null
+++ b/HuggingFaceTB-SmolLM2-135M-Instruct/CPU/requirements.txt
@@ -0,0 +1,3 @@
+olive-ai[auto-opt]
+onnxruntime>=1.20.1
+onnxruntime-genai>=0.5.0
\ No newline at end of file