diff --git a/.github/workflows/causal_lm_cpp.yml b/.github/workflows/causal_lm_cpp.yml
index 2e0afaa882..9cc60648ad 100644
--- a/.github/workflows/causal_lm_cpp.yml
+++ b/.github/workflows/causal_lm_cpp.yml
@@ -120,7 +120,10 @@ jobs:
           with open('pred.txt', 'r') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
-          tokenized = tokenizer('Why is the Sun yellow?', return_tensors='pt')
+          prompt = 'Why is the Sun yellow?'
+          if tokenizer.chat_template:
+              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
               ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref)
@@ -136,7 +139,10 @@ jobs:
           with open('pred.txt', 'r') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
-          tokenized = tokenizer('69', return_tensors='pt')
+          prompt = '69'
+          if tokenizer.chat_template:
+              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
               ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref)
@@ -152,7 +158,10 @@ jobs:
           with open('pred.txt', 'r') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
-          tokenized = tokenizer('Hi', return_tensors='pt')
+          prompt = 'Hi'
+          if tokenizer.chat_template:
+            prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
               ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref)
@@ -168,7 +177,10 @@ jobs:
           with open('pred.txt', 'r') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
-          tokenized = tokenizer('return 0', return_tensors='pt')
+          prompt = 'return 0'
+          if tokenizer.chat_template:
+              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
               ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref)
@@ -184,7 +196,10 @@ jobs:
           with open('pred.txt', 'r', errors='ignore') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
-          tokenized = tokenizer('你好！ 你好嗎？', return_tensors='pt')
+          prompt = '你好！ 你好嗎？'
+          if tokenizer.chat_template:
+              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
               ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref.replace('�', ''))
@@ -206,7 +221,9 @@ jobs:
             '你好！ 你好嗎？'
           ]
           for prompt in prompts:
-            tokenized = tokenizer(prompt, return_tensors='pt')
+            if tokenizer.chat_template:
+                prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+            tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
             for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
                 ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
                 idx = predictions.find(ref.replace('�', ''))
@@ -255,7 +272,10 @@ jobs:
           echo import transformers > ref.py
           echo predictions = open('cpp.txt', 'r').read() >> ref.py
           echo tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0', trust_remote_code=True) >> ref.py
-          echo tokenized = tokenizer('69', return_tensors='pt') >> ref.py
+          echo prompt = '69'
+          echo if tokenizer.chat_template:
+          echo     prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          echo tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False) >> ref.py
           echo for beam in transformers.AutoModelForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0', trust_remote_code=True).generate(**tokenized, max_new_tokens=100, do_sample=False): >> ref.py
           echo     ref = tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True) >> ref.py
           echo     idx = predictions.find(ref) >> ref.py
@@ -562,7 +582,10 @@ jobs:
           with open('pred_greedy.txt', 'r') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('microsoft/phi-1_5')
-          tokenized = tokenizer('Alan Turing was a', return_tensors='pt')
+          prompt = 'Alan Turing was a'
+          if tokenizer.chat_template:
+              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for output in transformers.AutoModelForCausalLM.from_pretrained('microsoft/phi-1_5').generate(**tokenized, max_length=100, do_sample=False):
               ref = tokenizer.decode(output[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref)
@@ -617,7 +640,10 @@ jobs:
           with open('pred_greedy.txt', 'r') as file:
               predictions = file.read()
           tokenizer = transformers.AutoTokenizer.from_pretrained('ikala/redpajama-3b-chat')
-          tokenized = tokenizer('Alan Turing was a', return_tensors='pt')
+          prompt = 'Alan Turing was a'
+          if tokenizer.chat_template:
+              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
+          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
           for output in transformers.AutoModelForCausalLM.from_pretrained('ikala/redpajama-3b-chat').generate(**tokenized, max_length=100, do_sample=False):
               ref = tokenizer.decode(output[tokenized['input_ids'].numel():], skip_special_tokens=True)
               idx = predictions.find(ref)
diff --git a/README.md b/README.md
index cea1e358bc..221a81c6c3 100644
--- a/README.md
+++ b/README.md
@@ -133,7 +133,6 @@ from PIL import Image
 
 # Choose GPU instead of CPU in the line below to run the model on Intel integrated or discrete GPU
 pipe = openvino_genai.VLMPipeline("./InternVL2-1B", "CPU")
-pipe.start_chat()
 
 image = Image.open("dog.jpg")
 image_data = np.array(image.getdata()).reshape(1, image.size[1], image.size[0], 3).astype(np.uint8)
diff --git a/samples/cpp/text_generation/README.md b/samples/cpp/text_generation/README.md
index d9e5bd8d22..c05fa25f9c 100644
--- a/samples/cpp/text_generation/README.md
+++ b/samples/cpp/text_generation/README.md
@@ -62,7 +62,7 @@ Recommended models: meta-llama/Llama-2-7b-chat-hf, TinyLlama/TinyLlama-1.1B-Chat
   ./chat_sample <MODEL_DIR>
   ```
 #### Missing chat template
-If you encounter an exception indicating a missing "chat template" when launching the `ov::genai::LLMPipeline` in chat mode, it likely means the model was not tuned for chat functionality. To work this around, manually add the chat template to tokenizer_config.json of your model.
+If you encounter an exception indicating a missing "chat template" when launching the `ov::genai::LLMPipeline` in chat mode, it likely means the model was not tuned for chat functionality. To work this around, manually add the chat template to tokenizer_config.json of your model or update it using call `pipe.get_tokenizer().set_chat_template(new_chat_template)`.
 The following template can be used as a default, but it may not work properly with every model:
 ```
 "chat_template": "{% for message in messages %}{% if (message['role'] == 'user') %}{{'<|im_start|>user\n' + message['content'] + '<|im_end|>\n<|im_start|>assistant\n'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|im_end|>\n'}}{% endif %}{% endfor %}",
diff --git a/samples/python/text_generation/README.md b/samples/python/text_generation/README.md
index 9940904cfb..db2f6b0d5f 100644
--- a/samples/python/text_generation/README.md
+++ b/samples/python/text_generation/README.md
@@ -62,7 +62,7 @@ Recommended models: meta-llama/Llama-2-7b-chat-hf, TinyLlama/TinyLlama-1.1B-Chat
   python chat_sample.py model_dir
   ```
 #### Missing chat template
-If you encounter an exception indicating a missing "chat template" when launching the `ov::genai::LLMPipeline` in chat mode, it likely means the model was not tuned for chat functionality. To work this around, manually add the chat template to tokenizer_config.json of your model.
+If you encounter an exception indicating a missing "chat template" when launching the `ov::genai::LLMPipeline` in chat mode, it likely means the model was not tuned for chat functionality. To work this around, manually add the chat template to tokenizer_config.json of your model or update it using call `pipe.get_tokenizer().set_chat_template(new_chat_template)`.
 The following template can be used as a default, but it may not work properly with every model:
 ```
 "chat_template": "{% for message in messages %}{% if (message['role'] == 'user') %}{{'<|im_start|>user\n' + message['content'] + '<|im_end|>\n<|im_start|>assistant\n'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|im_end|>\n'}}{% endif %}{% endfor %}",
diff --git a/src/README.md b/src/README.md
index af4953f98a..c2ed8c2a60 100644
--- a/src/README.md
+++ b/src/README.md
@@ -73,6 +73,8 @@ output:
 'it is made up of carbon atoms. The carbon atoms are arranged in a linear pattern, which gives the yellow color. The arrangement of carbon atoms in'
 ```
 
+>**Note**: The chat_template from tokenizer_config.json or from tokenizer/detokenizer model will be automatically applied to the prompt at the generation stage. If you want to disable it, you can do it by calling pipe.get_tokenizer().set_chat_template("").
+
 A simple chat in Python:
 ```python
 import openvino_genai as ov_genai
diff --git a/src/cpp/include/openvino/genai/generation_config.hpp b/src/cpp/include/openvino/genai/generation_config.hpp
index 3a75fc02ea..cd372e635d 100644
--- a/src/cpp/include/openvino/genai/generation_config.hpp
+++ b/src/cpp/include/openvino/genai/generation_config.hpp
@@ -128,6 +128,8 @@ class OPENVINO_GENAI_EXPORTS GenerationConfig {
 
     std::optional<AdapterConfig> adapters;
 
+    bool apply_chat_template = true;
+
     /** @brief sets eos_token_id to tokenizer_eos_token_id if eos_token_id is less than 0.
      * Otherwise verifies eos_token_id == tokenizer_eos_token_id.
      */
diff --git a/src/cpp/include/openvino/genai/llm_pipeline.hpp b/src/cpp/include/openvino/genai/llm_pipeline.hpp
index 31b1ac1675..26232574dc 100644
--- a/src/cpp/include/openvino/genai/llm_pipeline.hpp
+++ b/src/cpp/include/openvino/genai/llm_pipeline.hpp
@@ -177,6 +177,8 @@ class OPENVINO_GENAI_EXPORTS LLMPipeline {
     * @param generation_config optional GenerationConfig
     * @param streamer optional streamer
     * @return DecodedResults decoded resulting text
+    * chat_template will be applied to the prompt, run pipe.get_tokenizer().set_chat_template(custom_chat_template) to update it.
+    * To disable it for non-chat mode, please, use custom_chat_template eq "" or set generation_config.apply_chat_template to false.
     */
     DecodedResults generate(
         StringInputs inputs,
@@ -191,6 +193,8 @@ class OPENVINO_GENAI_EXPORTS LLMPipeline {
     * @param inputs input prompt or a vector of prompts
     * @param properties properties
     * @return DecodedResults decoded resulting text
+    * chat_template will be applied to the prompt, run pipe.get_tokenizer().set_chat_template(custom_chat_template) to update it.
+    * To disable it for non-chat mode, please, use custom_chat_template eq "" or set generation_config.apply_chat_template to false.
     */
     template <typename... Properties>
     util::EnableIfAllStringAny<DecodedResults, Properties...> generate(
diff --git a/src/cpp/include/openvino/genai/tokenizer.hpp b/src/cpp/include/openvino/genai/tokenizer.hpp
index 0a54d1da2a..bde4eb3fe1 100644
--- a/src/cpp/include/openvino/genai/tokenizer.hpp
+++ b/src/cpp/include/openvino/genai/tokenizer.hpp
@@ -221,6 +221,9 @@ class OPENVINO_GENAI_EXPORTS Tokenizer {
     /// @param chat_template The new template to override with.
     void set_chat_template(const std::string& chat_template);
 
+    // get information about a chat template to check its status, for example whether it is empty
+    std::string get_chat_template() const;
+
     // information about <bos>, <eos> tokens should be public,
     // they are used at least in StreamerBase descendants
     int64_t get_bos_token_id() const;
diff --git a/src/cpp/include/openvino/genai/visual_language/pipeline.hpp b/src/cpp/include/openvino/genai/visual_language/pipeline.hpp
index 8c3d380b0f..b6b1d5c7f6 100644
--- a/src/cpp/include/openvino/genai/visual_language/pipeline.hpp
+++ b/src/cpp/include/openvino/genai/visual_language/pipeline.hpp
@@ -98,6 +98,8 @@ class OPENVINO_GENAI_EXPORTS VLMPipeline {
     /// @param generation_config A config to follow for text generation.
     /// @param streamer A streamer to acquire intermediate result.
     /// @return A string generated by a model.
+    /// chat_template will be applied to the prompt, run pipe.set_chat_template(custom_chat_template) to update it.
+    /// To disable it for non-chat mode, please, use custom_chat_template eq "" or set generation_config.apply_chat_template to false.
     VLMDecodedResults generate(
         const std::string& prompt,
         const std::vector<ov::Tensor>& rgbs,
@@ -111,6 +113,8 @@ class OPENVINO_GENAI_EXPORTS VLMPipeline {
     /// @param generation_config A config to follow for text generation.
     /// @param streamer A streamer to acquire intermediate result.
     /// @return A string generated by a model.
+    /// chat_template will be applied to the prompt, run pipe.set_chat_template(custom_chat_template) to update it.
+    /// To disable it for non-chat mode, please, use custom_chat_template eq "" or set generation_config.apply_chat_template to false.
     VLMDecodedResults generate(
         const std::string& prompt,
         const ov::Tensor& rgb,
@@ -124,6 +128,8 @@ class OPENVINO_GENAI_EXPORTS VLMPipeline {
     /// for its members, StreamerVariant a single image or multiple
     /// images.
     /// @return A string generated by a model.
+    /// chat_template will be applied to the prompt, run pipe.set_chat_template(custom_chat_template) to update it.
+    /// To disable it for non-chat mode, please, use custom_chat_template eq "" or set generation_config.apply_chat_template to false.
     VLMDecodedResults generate(
         const std::string& prompt,
         const ov::AnyMap& config_map
@@ -137,6 +143,8 @@ class OPENVINO_GENAI_EXPORTS VLMPipeline {
     /// @param ...properties ov::Property instances to be combined into
     /// ov::AnyMap.
     /// @return A string generated by a model.
+    /// chat_template will be applied to the prompt, run pipe.set_chat_template(custom_chat_template) to update it.
+    /// To disable it for non-chat mode, please, use custom_chat_template eq "" or set generation_config.apply_chat_template to false.
     template <typename... Properties>
     util::EnableIfAllStringAny<VLMDecodedResults, Properties...> generate(
         const std::string& prompt,
diff --git a/src/cpp/src/generation_config.cpp b/src/cpp/src/generation_config.cpp
index 67682be787..dba5aaa5bd 100644
--- a/src/cpp/src/generation_config.cpp
+++ b/src/cpp/src/generation_config.cpp
@@ -125,6 +125,7 @@ void GenerationConfig::update_generation_config(const ov::AnyMap& properties) {
     read_anymap_param(properties, "logprobs", logprobs);
     read_anymap_param(properties, "num_return_sequences", num_return_sequences);
     read_anymap_param(properties, "adapters", adapters);
+    read_anymap_param(properties, "apply_chat_template", apply_chat_template);
 
     // penalties
     read_anymap_param(properties, "frequency_penalty", frequency_penalty);
diff --git a/src/cpp/src/icontinuous_batching.cpp b/src/cpp/src/icontinuous_batching.cpp
index 78f8fda8f7..6b748d6665 100644
--- a/src/cpp/src/icontinuous_batching.cpp
+++ b/src/cpp/src/icontinuous_batching.cpp
@@ -53,9 +53,20 @@ ContinuousBatchingPipeline::IContinuousBatchingPipeline::generate(
     } else {
         input_ids.reserve(prompts.size());
         timer.start();
-        for (const std::string& prompt : prompts) {
+        for (size_t i = 0; i < prompts.size(); i++) {
+            const std::string& prompt = prompts.at(i);
             const auto encode_start = std::chrono::steady_clock::now();
-            input_ids.push_back(m_tokenizer.encode(prompt).input_ids);
+            ov::Tensor encoded_inputs;
+            if (sampling_params.at(i).apply_chat_template && !m_tokenizer.get_chat_template().empty()) {
+                ChatHistory history({{{"role", "user"}, {"content", prompt}}});
+                constexpr bool add_generation_prompt = true;
+                auto templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt);
+                encoded_inputs = m_tokenizer.encode(templated_prompt, ov::genai::add_special_tokens(false)).input_ids;
+            } else {
+                // in case when chat_template was not found in tokenizer_config.json or set
+                encoded_inputs = m_tokenizer.encode(prompt).input_ids;
+            }
+            input_ids.push_back(encoded_inputs);
             tokenization_durations.emplace_back(PerfMetrics::get_microsec(std::chrono::steady_clock::now() - encode_start));
         }
         timer.end();
diff --git a/src/cpp/src/llm_pipeline_stateful.cpp b/src/cpp/src/llm_pipeline_stateful.cpp
index 2a53154c27..14a3f1b651 100644
--- a/src/cpp/src/llm_pipeline_stateful.cpp
+++ b/src/cpp/src/llm_pipeline_stateful.cpp
@@ -88,7 +88,19 @@ DecodedResults StatefulLLMPipeline::generate(
 
     if (auto input_vector = std::get_if<std::vector<std::string>>(&inputs)) {
         OPENVINO_ASSERT(!is_chat_conversation, "Can't chat with multiple prompts");
-        encoded_input = m_tokenizer.encode(*input_vector);
+        std::vector<std::string> templated_input_vector;
+        for (auto& input : *input_vector) {
+            if (config.apply_chat_template && !m_tokenizer.get_chat_template().empty()) {
+                ChatHistory history({{{"role", "user"}, {"content", input}}});
+                constexpr bool add_generation_prompt = true;
+                auto templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt);
+                templated_input_vector.push_back(templated_prompt);
+            } else {
+                // in case when chat_template was not found in tokenizer_config.json or set
+                templated_input_vector.push_back(input);
+            }
+        }
+        encoded_input = m_tokenizer.encode(templated_input_vector, ov::genai::add_special_tokens(false));
     } else if (auto input_prompt = std::get_if<std::string>(&inputs)) {
         std::string& prompt = *input_prompt;
 
@@ -104,7 +116,7 @@ DecodedResults StatefulLLMPipeline::generate(
 
             m_history.push_back({{"role", "user"}, {"content", prompt}});
             constexpr bool add_generation_prompt = true;
-            auto new_templated_chat_history  = m_tokenizer.apply_chat_template(m_history, add_generation_prompt);
+            auto new_templated_chat_history = m_tokenizer.apply_chat_template(m_history, add_generation_prompt);
             // Do not add special tokens in chat scenario to be aligned with HF.
             auto new_chat_tokens = m_tokenizer.encode(new_templated_chat_history, ov::genai::add_special_tokens(false));
             auto prev_chat_tokens = m_tokenizer.encode(m_templated_chat_history, ov::genai::add_special_tokens(false));
@@ -157,7 +169,16 @@ DecodedResults StatefulLLMPipeline::generate(
 
             // TODO: Forbid LoRA config change if we are in the chat mode, because it requires regenerating the history with LoRA applied
         } else {
-            encoded_input = m_tokenizer.encode(prompt);
+            std::string& prompt = *input_prompt;
+            if (config.apply_chat_template && !m_tokenizer.get_chat_template().empty()) {
+                ChatHistory history({{{"role", "user"}, {"content", prompt}}});
+                constexpr bool add_generation_prompt = true;
+                auto templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt);
+                encoded_input = m_tokenizer.encode(templated_prompt, ov::genai::add_special_tokens(false));
+            } else {
+                // in case when chat_template was not found in tokenizer_config.json or set
+                encoded_input = m_tokenizer.encode(prompt);
+            }
         }
     }
 
diff --git a/src/cpp/src/llm_pipeline_static.cpp b/src/cpp/src/llm_pipeline_static.cpp
index b9a55477cb..1f4e5ffcec 100644
--- a/src/cpp/src/llm_pipeline_static.cpp
+++ b/src/cpp/src/llm_pipeline_static.cpp
@@ -805,7 +805,15 @@ DecodedResults StatefulLLMPipeline::generate(
         // for chat ov::genai::add_special_tokens(false) is aligned with stateful pipeline and HF
         tokenized_input = m_tokenizer.encode(prompt, ov::genai::add_special_tokens(false));
     } else {
-        tokenized_input = m_tokenizer.encode(prompt);
+        if (config.apply_chat_template && !m_tokenizer.get_chat_template().empty()) {
+            ChatHistory history({{{"role", "user"}, {"content", prompt}}});
+            constexpr bool add_generation_prompt = true;
+            auto templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt);
+            tokenized_input = m_tokenizer.encode(templated_prompt, ov::genai::add_special_tokens(false));
+        } else {
+            // in case when chat_template was not found in tokenizer_config.json or set
+            tokenized_input = m_tokenizer.encode(prompt);
+        }
     }
 
     auto encode_stop_time =  std::chrono::steady_clock::now();
@@ -1273,7 +1281,15 @@ DecodedResults StatelessLLMPipeline::generate(
         // for chat ov::genai::add_special_tokens(false) is aligned with stateful pipeline and HF
         tokenized_input = m_tokenizer.encode(prompt, ov::genai::add_special_tokens(false));
     } else {
-        tokenized_input = m_tokenizer.encode(prompt);
+        if (config.apply_chat_template && !m_tokenizer.get_chat_template().empty()) {
+            ChatHistory history({{{"role", "user"}, {"content", prompt}}});
+            constexpr bool add_generation_prompt = true;
+            auto templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt);
+            tokenized_input = m_tokenizer.encode(templated_prompt, ov::genai::add_special_tokens(false));
+        } else {
+            // in case when chat_template was not found in tokenizer_config.json or set
+            tokenized_input = m_tokenizer.encode(prompt);
+        }
     }
 
     auto encode_stop_time =  std::chrono::steady_clock::now();
diff --git a/src/cpp/src/tokenizer.cpp b/src/cpp/src/tokenizer.cpp
index 2f18e87839..441b422584 100644
--- a/src/cpp/src/tokenizer.cpp
+++ b/src/cpp/src/tokenizer.cpp
@@ -587,6 +587,10 @@ class Tokenizer::TokenizerImpl {
     void set_chat_template(const std::string& chat_template) {
         m_chat_template = patch_chat_template(chat_template);
     }
+
+    std::string get_chat_template() {
+        return m_chat_template;
+    }
 };
 
 Tokenizer::Tokenizer(const std::filesystem::path& tokenizer_path, const ov::AnyMap& properties) {
@@ -690,6 +694,10 @@ std::string Tokenizer::apply_chat_template(ChatHistory history,
     return m_pimpl->apply_chat_template(history, add_generation_prompt, chat_template);
 }
 
+std::string Tokenizer::get_chat_template() const {
+    return m_pimpl->get_chat_template();
+}
+
 void Tokenizer::set_chat_template(const std::string& chat_template) {
     m_pimpl->set_chat_template(chat_template);
 }
diff --git a/src/cpp/src/visual_language/inputs_embedder.cpp b/src/cpp/src/visual_language/inputs_embedder.cpp
index 4f3812862c..9608bce385 100644
--- a/src/cpp/src/visual_language/inputs_embedder.cpp
+++ b/src/cpp/src/visual_language/inputs_embedder.cpp
@@ -50,7 +50,7 @@ class InputsEmbedder::IInputsEmbedder {
     ov::genai::utils::HistoryRemoveManager m_kv_history_manager = {0, 0};
 
 public:
-    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) = 0;
+    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) = 0;
 
     virtual std::pair<ov::Tensor, std::optional<int64_t>> get_position_ids(const size_t inputs_embeds_size, const size_t history_size) {
         ov::Tensor position_ids = ov::Tensor{ov::element::i64, { 1, inputs_embeds_size }};
@@ -155,7 +155,7 @@ class InputsEmbedder::IInputsEmbedder {
         ),
         m_tokenizer(tokenizer) { }
 
-    ov::Tensor get_encoded_input_ids(const std::string& prompt, ov::genai::VLMPerfMetrics& metrics, const std::string& chat_template_fallback = {}) {
+    ov::Tensor get_encoded_input_ids(const std::string& prompt, ov::genai::VLMPerfMetrics& metrics, const std::string& chat_template_fallback = {}, bool apply_chat_template = true) {
         ov::Tensor encoded_input_ids;
         if (m_is_chat_conversation) {
             // KV cache in model already contains prompts and answers from previous iterations.
@@ -169,9 +169,9 @@ class InputsEmbedder::IInputsEmbedder {
             m_history.push_back({{"role", "user"}, {"content", prompt}});
             constexpr bool add_generation_prompt = true;
             std::string new_templated_chat_history;
-            try {
+            if (!m_tokenizer.get_chat_template().empty()) {
                 new_templated_chat_history = m_tokenizer.apply_chat_template(m_history, add_generation_prompt);
-            } catch (const std::exception& error) {
+            } else {
                 // Use fallback chat template if it was not found in tokenizer_config.json
                 new_templated_chat_history = m_tokenizer.apply_chat_template(m_history, add_generation_prompt, chat_template_fallback);
             }
@@ -224,7 +224,21 @@ class InputsEmbedder::IInputsEmbedder {
             std::copy_n(new_chat_tokens.data<int64_t>(), new_chat_tokens.get_size(), std::back_inserter(m_tokenized_history));
         } else {
             auto start_tokenizer_time = std::chrono::steady_clock::now();
-            encoded_input_ids = m_tokenizer.encode(prompt).input_ids;
+            if (apply_chat_template) {
+                std::string templated_prompt;
+                ChatHistory history({{{"role", "user"}, {"content", prompt}}});
+                constexpr bool add_generation_prompt = true;
+
+                if (!m_tokenizer.get_chat_template().empty()) {
+                    templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt);
+                } else {
+                    // Use fallback chat template if it was not found in tokenizer_config.json
+                    templated_prompt = m_tokenizer.apply_chat_template(history, add_generation_prompt, chat_template_fallback);
+                }
+                encoded_input_ids = m_tokenizer.encode(templated_prompt, ov::genai::add_special_tokens(false)).input_ids;
+            } else {
+                encoded_input_ids = m_tokenizer.encode(prompt).input_ids;
+            }
             auto end_tokenizer_time = std::chrono::steady_clock::now();
             metrics.raw_metrics.tokenization_durations.emplace_back(PerfMetrics::get_microsec(end_tokenizer_time - start_tokenizer_time));
             m_tokenized_history.clear();
@@ -320,7 +334,7 @@ class InputsEmbedderMiniCPM : public InputsEmbedder::IInputsEmbedder {
             m_pos_embed_cache = get_2d_sincos_pos_embed(m_vlm_config.hidden_size, {70, 70});
         }
 
-    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) override {
+    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) override {
         std::string images_prompt;
         std::vector<EncodedImage> embeds;
 
@@ -355,7 +369,7 @@ class InputsEmbedderMiniCPM : public InputsEmbedder::IInputsEmbedder {
         }
         images_prompt += prompt;
 
-        ov::Tensor encoded_input = get_encoded_input_ids(images_prompt, metrics);
+        ov::Tensor encoded_input = get_encoded_input_ids(images_prompt, metrics, {}, apply_chat_template);
 
         ov::Tensor inputs_embeds = m_embedding.infer(encoded_input);
         OPENVINO_ASSERT(
@@ -618,7 +632,7 @@ class InputsEmbedderLLaVA : public InputsEmbedder::IInputsEmbedder {
         const ov::AnyMap device_config) :
         IInputsEmbedder(vlm_config, models_map, tokenizer, config_dir_path, device, device_config) { }
 
-    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) override {
+    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) override {
         std::string image_token = m_vlm_config.im_start;
         // Adapted from llava-1.5-7b-hf chat_template.json
         std::string chat_template_fallback = "{% for message in messages %}{% if message['role'] == 'user' %}{{ 'USER: ' + message['content'] + ' ' }}{% else %}{{ 'ASSISTANT: ' + message['content'] + ' ' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ 'ASSISTANT:' }}{% endif %}";
@@ -636,7 +650,7 @@ class InputsEmbedderLLaVA : public InputsEmbedder::IInputsEmbedder {
         }
         formatted_prompt += prompt;
 
-        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, chat_template_fallback);
+        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, chat_template_fallback, apply_chat_template);
         ov::Tensor text_embeds = m_embedding.infer(input_ids);
 
         if (images.empty()) {
@@ -731,7 +745,7 @@ class InputsEmbedderLLaVANext : public InputsEmbedderLLaVA {
         const ov::AnyMap device_config) :
         InputsEmbedderLLaVA(vlm_config, models_map, tokenizer, config_dir_path, device, device_config) { }
 
-    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) override {
+    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) override {
         std::string image_token = m_vlm_config.im_start;
         // Adapted from llava-1.5-7b-hf chat_template.json
         std::string chat_template_fallback = "{% for message in messages %}{% if message['role'] == 'user' %}{{ 'USER: ' + message['content'] + ' ' }}{% else %}{{ 'ASSISTANT: ' + message['content'] + ' ' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ 'ASSISTANT:' }}{% endif %}";
@@ -763,7 +777,7 @@ class InputsEmbedderLLaVANext : public InputsEmbedderLLaVA {
         }
         formatted_prompt += prompt;
 
-        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, chat_template_fallback);
+        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, chat_template_fallback, apply_chat_template);
         ov::Tensor text_embeds = m_embedding.infer(input_ids);
 
         if (images.empty()) {
@@ -1058,7 +1072,7 @@ class InputsEmbedderInternVLChat : public InputsEmbedder::IInputsEmbedder {
         const ov::AnyMap device_config) :
         IInputsEmbedder(vlm_config, models_map, tokenizer, config_dir_path, device, device_config) { }
 
-    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) override {
+    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) override {
         std::string image_start_token = m_vlm_config.image_start_token;
         std::string image_context_token = m_vlm_config.image_context_token;
         std::string image_end_token = m_vlm_config.image_end_token;
@@ -1086,7 +1100,7 @@ class InputsEmbedderInternVLChat : public InputsEmbedder::IInputsEmbedder {
         }
         formatted_prompt += prompt;
 
-        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics);
+        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, {}, apply_chat_template);
         ov::Tensor text_embeds = m_embedding.infer(input_ids);
 
         if (images.empty()) {
@@ -1203,7 +1217,7 @@ class InputsEmbedderQwen2VL : public InputsEmbedder::IInputsEmbedder {
             ).create_infer_request();
         }
     
-    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) override {
+    virtual ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) override {
         std::string formatted_prompt;
 
         std::vector<ov::Tensor> single_images = to_single_image_tensors(images);
@@ -1235,7 +1249,7 @@ class InputsEmbedderQwen2VL : public InputsEmbedder::IInputsEmbedder {
 
         // Adapted from Qwen/Qwen2-7B-Instruct
         std::string chat_template_fallback = "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}";
-        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, chat_template_fallback);
+        ov::Tensor input_ids = get_encoded_input_ids(formatted_prompt, metrics, chat_template_fallback, apply_chat_template);
         ov::Tensor text_embeds = m_embedding.infer(input_ids);
 
         if (images.empty()) {
@@ -1605,8 +1619,8 @@ InputsEmbedder::InputsEmbedder(const VLMConfig& vlm_config,
     }
 }
 
-ov::Tensor InputsEmbedder::get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics) {
-    return m_impl->get_inputs_embeds(prompt, images, metrics);
+ov::Tensor InputsEmbedder::get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template) {
+    return m_impl->get_inputs_embeds(prompt, images, metrics, apply_chat_template);
 }
 
 std::pair<ov::Tensor, std::optional<int64_t>> InputsEmbedder::get_position_ids(const size_t inputs_embeds_size, const size_t history_size) {
diff --git a/src/cpp/src/visual_language/inputs_embedder.hpp b/src/cpp/src/visual_language/inputs_embedder.hpp
index 223d090b22..b21d0198cc 100644
--- a/src/cpp/src/visual_language/inputs_embedder.hpp
+++ b/src/cpp/src/visual_language/inputs_embedder.hpp
@@ -32,7 +32,7 @@ class InputsEmbedder {
                    const ov::AnyMap device_config);
 
     // compute input embedding for prompt and multiple images
-    ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics);
+    ov::Tensor get_inputs_embeds(const std::string& prompt, const std::vector<ov::Tensor>& images, ov::genai::VLMPerfMetrics& metrics, bool apply_chat_template);
 
     // compute position ids for language model input
     std::pair<ov::Tensor, std::optional<int64_t>> get_position_ids(const size_t inputs_embeds_size, const size_t history_size);
diff --git a/src/cpp/src/visual_language/pipeline.cpp b/src/cpp/src/visual_language/pipeline.cpp
index 95e3064548..415e7fcae9 100644
--- a/src/cpp/src/visual_language/pipeline.cpp
+++ b/src/cpp/src/visual_language/pipeline.cpp
@@ -166,7 +166,7 @@ class ov::genai::VLMPipeline::VLMPipelineImpl {
         generation_config.validate();
 
         auto start_get_inputs_embeds = std::chrono::steady_clock::now();
-        ov::Tensor inputs_embeds = m_inputs_embedder->get_inputs_embeds(prompt, rgbs, perf_metrics);
+        ov::Tensor inputs_embeds = m_inputs_embedder->get_inputs_embeds(prompt, rgbs, perf_metrics, generation_config.apply_chat_template);
         auto end_get_inputs_embeds = std::chrono::steady_clock::now();
 
         auto to_remove_from_hist = m_inputs_embedder->get_num_tokens_to_remove_from_hist();
diff --git a/src/python/py_generation_config.cpp b/src/python/py_generation_config.cpp
index e2a6d7062c..a7d7789a55 100644
--- a/src/python/py_generation_config.cpp
+++ b/src/python/py_generation_config.cpp
@@ -115,6 +115,7 @@ void init_generation_config(py::module_& m) {
         .def_readwrite("include_stop_str_in_output", &GenerationConfig::include_stop_str_in_output)
         .def_readwrite("stop_token_ids", &GenerationConfig::stop_token_ids)
         .def_readwrite("adapters", &GenerationConfig::adapters)
+        .def_readwrite("apply_chat_template", &GenerationConfig::apply_chat_template)
         .def("set_eos_token_id", &GenerationConfig::set_eos_token_id, py::arg("tokenizer_eos_token_id"))
         .def("is_beam_search", &GenerationConfig::is_beam_search)
         .def("is_greedy_decoding", &GenerationConfig::is_greedy_decoding)
diff --git a/tests/python_tests/common.py b/tests/python_tests/common.py
index cbed01e20f..8f2d2b05b9 100644
--- a/tests/python_tests/common.py
+++ b/tests/python_tests/common.py
@@ -251,6 +251,8 @@ def run_hugging_face(
         # process prompt by promp as we have multiple generation configs
         for prompt, generation_config in zip(prompts, generation_configs):
             hf_generation_config = convert_to_hf(opt_model.generation_config, generation_config)
+            if hf_tokenizer.chat_template:
+                prompt = hf_tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
             inputs = hf_tokenizer(prompt, return_tensors="pt")
             input_ids, attention_mask = inputs['input_ids'], inputs['attention_mask']
             prompt_len = 0 if generation_config.echo else input_ids.numel()
@@ -265,8 +267,14 @@ def run_hugging_face(
                 generation_result.m_scores = [score for score in generate_outputs.sequences_scores]
             generation_results.append(generation_result)
     else:
+        processed_prompts = []
+        if hf_tokenizer.chat_template:
+            for prompt in prompts:
+                processed_prompts.append(hf_tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True))
+        else:
+            processed_prompts = prompts
         # process all prompts as a single batch as we have a single generation config for all prompts
-        inputs = hf_tokenizer(prompts, return_tensors='pt', padding=True, truncation=True, add_special_tokens=True, padding_side='left')
+        inputs = hf_tokenizer(processed_prompts, return_tensors='pt', padding=True, truncation=True, add_special_tokens=False, padding_side='left')
         input_ids, attention_mask = inputs['input_ids'], inputs['attention_mask']
         hf_generation_config = convert_to_hf(opt_model.generation_config, generation_configs)
         hf_encoded_outputs = opt_model.generate(input_ids, attention_mask=attention_mask, generation_config=hf_generation_config, tokenizer=hf_tokenizer)