diff --git a/examples/training/distillation/model_distillation.py b/examples/training/distillation/model_distillation.py
index b3b6f69dd..9f89af3d2 100644
--- a/examples/training/distillation/model_distillation.py
+++ b/examples/training/distillation/model_distillation.py
@@ -56,16 +56,6 @@
 inference_batch_size = 64
 train_batch_size = 64
 
-# We use AllNLI as a source of sentences for the distillation
-nli_dataset_path = "datasets/AllNLI.tsv.gz"
-
-# Further, we use sentences extracted from the English Wikipedia to train the distillation
-wikipedia_dataset_path = "datasets/wikipedia-en-sentences.txt.gz"
-
-# We use the STS benchmark dataset to see how much performance we loose
-sts_dataset_path = "datasets/stsbenchmark.tsv.gz"
-
-
 logging.info("Load the AllNLI dataset")
 # Load the AllNLI dataset: https://huggingface.co/datasets/sentence-transformers/all-nli
 nli_train_dataset = load_dataset("sentence-transformers/all-nli", "pair-score", split="train")
diff --git a/examples/training/distillation/model_quantization.py b/examples/training/distillation/model_quantization.py
index f0d621fc7..08427d03a 100644
--- a/examples/training/distillation/model_quantization.py
+++ b/examples/training/distillation/model_quantization.py
@@ -1,93 +1,158 @@
 """
 A quantized model executes some or all of the operations with integers rather than floating point values. This allows for a more compact models and the use of high performance vectorized operations on many hardware platforms.
 
-As a result, you get about 40% smaller and faster models. The speed-up depends on your CPU and how PyTorch was build and can be anywhere between 10% speed-up and 300% speed-up.
+As a result, you get about much smaller and faster models. The speed-up depends on your CPU, but you can expect a speed-up of 2x to 4x for most CPUs. The model size is also reduced by 2x.
 
-Note: Quantized models are only available for CPUs. Use a GPU, if available, for optimal performance.
+Note: Quantized models are only recommended for CPUs. If available, Use a GPU for optimal performance.
 
-For more details:
-https://pytorch.org/docs/stable/quantization.html
+See docs for more information on quantization, optimization, benchmarks, etc.: https://sbert.net/docs/sentence_transformer/usage/efficiency.html
 """
 
-import csv
-import gzip
 import logging
-import os
 import time
 
-import torch
-from torch.nn import Embedding, Linear
-from torch.quantization import quantize_dynamic
+from datasets import load_dataset
 
-from sentence_transformers import InputExample, LoggingHandler, SentenceTransformer, util
-from sentence_transformers.evaluation import EmbeddingSimilarityEvaluator
-
-#### Just some code to print debug information to stdout
-logging.basicConfig(
-    format="%(asctime)s - %(message)s", datefmt="%Y-%m-%d %H:%M:%S", level=logging.INFO, handlers=[LoggingHandler()]
+from sentence_transformers import (
+    SentenceTransformer,
+    export_dynamic_quantized_onnx_model,
+    export_static_quantized_openvino_model,
 )
-#### /print debug information to stdout
-
-
-# Check if dataset exists. If not, download and extract  it
-sts_dataset_path = "datasets/stsbenchmark.tsv.gz"
-
-if not os.path.exists(sts_dataset_path):
-    util.http_get("https://sbert.net/datasets/stsbenchmark.tsv.gz", sts_dataset_path)
-
-# Limit torch to 4 threads
-torch.set_num_threads(4)
+from sentence_transformers.evaluation import EmbeddingSimilarityEvaluator
 
 #### Just some code to print debug information to stdout
 logging.basicConfig(format="%(asctime)s - %(message)s", datefmt="%Y-%m-%d %H:%M:%S", level=logging.INFO)
-### /print debug information to stdout
 
-model_name = "all-distilroberta-v1"
-
-# Load a named sentence model (based on BERT). This will download the model from our server.
-# Alternatively, you can also pass a filepath to SentenceTransformer()
-model = SentenceTransformer(model_name, device="cpu")
-q_model = quantize_dynamic(model, {Linear, Embedding})
+# Load some sentences from the STSbenchmark dataset
+train_dataset = load_dataset("sentence-transformers/stsb", split="train")
+sentences = train_dataset["sentence1"] + train_dataset["sentence2"]
+sentences = sentences[:10_000]
+test_dataset = load_dataset("sentence-transformers/stsb", split="test")
 
+model_name = "all-mpnet-base-v2"
 
-# Convert the dataset to a DataLoader ready for training
-logging.info("Read STSbenchmark dataset")
-test_samples = []
-sentences = []
+# 1. Load a baseline model with just fp32 torch
+model = SentenceTransformer(model_name, device="cpu")
 
-with gzip.open(sts_dataset_path, "rt", encoding="utf8") as fIn:
-    reader = csv.DictReader(fIn, delimiter="\t", quoting=csv.QUOTE_NONE)
-    for row in reader:
-        score = float(row["score"]) / 5.0  # Normalize score to range 0 ... 1
-        inp_example = InputExample(texts=[row["sentence1"], row["sentence2"]], label=score)
+# 2. Load an ONNX model to quantize
+onnx_model = SentenceTransformer(
+    model_name,
+    backend="onnx",
+    device="cpu",
+    model_kwargs={"provider": "CPUExecutionProvider"},
+)
 
-        sentences.append(row["sentence1"])
-        sentences.append(row["sentence2"])
+# 3. Quantize the ONNX model
+quantized_onnx_model_path = f"{model_name.replace('/', '-')}-onnx-quantized"
+onnx_model.save_pretrained(quantized_onnx_model_path)
+export_dynamic_quantized_onnx_model(
+    onnx_model,
+    quantization_config="avx512_vnni",
+    model_name_or_path=quantized_onnx_model_path,
+)
+quantized_onnx_model = SentenceTransformer(
+    quantized_onnx_model_path,
+    backend="onnx",
+    device="cpu",
+    model_kwargs={
+        "file_name": "model_qint8_avx512_vnni.onnx",
+        "provider": "CPUExecutionProvider",
+    },
+)
+# Alternatively, you can load the pre-quantized model:
+# quantized_onnx_model = SentenceTransformer(
+#     model_name,
+#     backend="onnx",
+#     device="cpu",
+#     model_kwargs={
+#         "file_name": "model_qint8_avx512_vnni.onnx",
+#         "provider": "CPUExecutionProvider",
+#     },
+# )
+
+# To make sure that `onnx_model` itself didn't get quantized, we reload it
+onnx_model = SentenceTransformer(
+    model_name,
+    backend="onnx",
+    device="cpu",
+    model_kwargs={"provider": "CPUExecutionProvider"},
+)
 
-        if row["split"] == "test":
-            test_samples.append(inp_example)
+# 4. Load an OpenVINO model to quantize
+openvino_model = SentenceTransformer(model_name, backend="openvino", device="cpu")
 
-sentences = sentences[0:10000]
+# 5. Quantize the OpenVINO model
+quantized_ov_model_path = f"{model_name.replace('/', '-')}-ov-quantized"
+openvino_model.save_pretrained(quantized_ov_model_path)
+export_static_quantized_openvino_model(
+    openvino_model,
+    quantization_config=None,
+    model_name_or_path=quantized_ov_model_path,
+)
+quantized_ov_model = SentenceTransformer(
+    quantized_ov_model_path,
+    backend="openvino",
+    device="cpu",
+    model_kwargs={"file_name": "openvino_model_qint8_quantized.xml"},
+)
+# Alternatively, you can load the pre-quantized model:
+# quantized_ov_model = SentenceTransformer(
+#     model_name,
+#     backend="openvino",
+#     device="cpu",
+#     model_kwargs={"file_name": "openvino_model_qint8_quantized.xml"},
+# )
+
+# To make sure that `openvino_model` itself didn't get quantized, we reload it
+openvino_model = SentenceTransformer(model_name, backend="openvino", device="cpu")
+
+
+# Create a function to evaluate the models
+def evaluate(model: SentenceTransformer, name: str) -> None:
+    logging.info(f"Evaluating {name}")
+    start_time = time.time()
+    model.encode(sentences)
+    diff = time.time() - start_time
+    logging.info(f"Done after {diff:.2f} sec. {len(sentences) / diff:.2f} sentences / sec")
+
+    evaluator = EmbeddingSimilarityEvaluator(
+        sentences1=test_dataset["sentence1"],
+        sentences2=test_dataset["sentence2"],
+        scores=test_dataset["score"],
+        name="sts-test",
+    )
+    results = evaluator(model)
+    logging.info(f"STS Benchmark, {evaluator.primary_metric}: {results[evaluator.primary_metric]}")
+
+
+# Evaluate the models
+for model, name in [
+    (model, "Baseline"),
+    (onnx_model, "ONNX"),
+    (quantized_onnx_model, "Quantized ONNX"),
+    (openvino_model, "OpenVINO"),
+    (quantized_ov_model, "Quantized OpenVINO"),
+]:
+    evaluate(model, name)
 
-logging.info("Evaluating speed of unquantized model")
-start_time = time.time()
-emb = model.encode(sentences, show_progress_bar=True)
-diff_normal = time.time() - start_time
-logging.info(f"Done after {diff_normal:.2f} sec. {len(sentences) / diff_normal:.2f} sentences / sec")
+"""
+Evaluating Baseline
+Done after 48.79 sec. 204.97 sentences / sec
+STS Benchmark, sts-test_spearman_cosine: 0.834221557992808
 
-logging.info("Evaluating speed of quantized model")
-start_time = time.time()
-emb = q_model.encode(sentences, show_progress_bar=True)
-diff_quantized = time.time() - start_time
-logging.info(f"Done after {diff_quantized:.2f} sec. {len(sentences) / diff_quantized:.2f} sentences / sec")
-logging.info(f"Speed-up: {diff_normal / diff_quantized:.2f}")
-#########
+Evaluating ONNX
+Done after 36.79 sec. 271.84 sentences / sec
+STS Benchmark, sts-test_spearman_cosine: 0.8342216139244768
 
-evaluator = EmbeddingSimilarityEvaluator.from_input_examples(test_samples, name="sts-test")
+Evaluating Quantized ONNX
+Done after 17.84 sec. 560.60 sentences / sec
+STS Benchmark, sts-test_spearman_cosine: 0.8256725903061843
 
-logging.info("Evaluate regular model")
-model.evaluate(evaluator)
+Evaluating OpenVINO
+Done after 36.43 sec. 274.49 sentences / sec
+STS Benchmark, sts-test_spearman_cosine: 0.834221557992808
 
-print("\n\n")
-logging.info("Evaluate quantized model")
-q_model.evaluate(evaluator)
+Evaluating Quantized OpenVINO
+Done after 12.94 sec. 772.83 sentences / sec
+STS Benchmark, sts-test_spearman_cosine: 0.8315710087348848
+"""