@@ -38,6 +38,7 @@ from ml_model.services.perplexity import Perplexity
 from ml_model.services.pulid import Pulid
 from ml_model.services.qwen import Qwen
 from ml_model.services.qwen_235B import Qwen_235B
+from ml_model.services.qwen_3_max_thinking import Qwen_3_Max_Thinking
 from ml_model.services.raifgpt import Raifgpt
 from ml_model.services.ray import Ray
 from ml_model.services.recraft import Recraft
@@ -0,0 +1,86 @@
+import time
+from datetime import timedelta
+from decimal import Decimal
+
+from messages.models import Message
+from ml_model.services.base import SimpleService
+from ml_model.tasks import openrouter_run
+from tools.chats.models import Chat
+from tools.copywrite.models import Copywrite
+from tools.public_api.models import APIStore
+
+
+class Qwen_3_Max_Thinking(SimpleService):
+    TOKENS_COST = {
+        'input': {'default': Decimal('360'), 'high': Decimal('900')},
+        'output': {'default': Decimal('1800'), 'high': Decimal('4500')}
+    }
+
+    def calculate_price(self, input_tokens: int, output_tokens: int) -> Decimal:
+        price = (
+            input_tokens * (self.TOKENS_COST['input']['default' if input_tokens <= 128_000 else 'high'] / 1_000_000)
+            + output_tokens * (self.TOKENS_COST['output']['default' if input_tokens <= 128_000 else 'high'] / 1_000_000)
+            + Decimal('6')
+        )
+        return price.quantize(Decimal('0.1'), rounding='ROUND_UP')
+
+    def save_results(self, content: str, time: timedelta, save: bool = True) -> list[Message]:
+        msgs = [
+            Message(
+                content=content,
+                content_object=self.store,
+                elapsed_time=time,
+            )
+        ]
+        if save:
+            return Message.objects.bulk_create(msgs)
+        return msgs
+
+    def make(self, input_message: Message, save: bool = True) -> list[Message]:
+        start_time = time.time()
+        callback_data = {'provider': {'order': ['alibaba']}, **input_message.info}
+        messages = self.get_chat_history()
+        messages.append({'role': 'user', 'content': input_message.content})
+        result = openrouter_run('qwen/qwen3-max:online', messages, callback_data, 'Qwen')
+        process_time = timedelta(seconds=(time.time() - start_time))
+        self.handle_invoice(
+            input_message.content_object.model,
+            input_tokens=result[1],
+            output_tokens=result[2],
+        )
+        msgs = self.save_results(result[0], process_time)
+        return msgs
+
+    def get_chat_history(self, message_limit: int = 10, max_character_limit: int = 1500) -> list[dict[str, str | list]]:
+        if isinstance(self.store, Chat):
+            air_messages = list(
+                reversed(
+                    Message.objects.filter(
+                        chats_chats_messages=self.store, is_deleted=False, is_sent=True
+                    ).order_by('-created_at')[1:message_limit+1]
+                )
+            )
+        elif isinstance(self.store, APIStore):
+            air_messages = []
+        elif isinstance(self.store, Copywrite):
+            air_messages = list(
+                reversed(
+                    Message.objects.filter(
+                        copywrite_copywrites_messages=self.store,
+                        is_deleted=False,
+                        is_sent=True,
+                    ).order_by('-created_at')[:message_limit]
+                )
+            )
+        memory = []
+        for msg in air_messages:
+            content = msg.content or ''
+            if msg.from_model:
+                memory.append({'role': 'assistant', 'content': content})
+            else:
+                memory.append({'role': 'user', 'content': content})
+        character_length = sum(len(content['content']) for content in memory)
+        while character_length > max_character_limit:
+            character_length -= len(memory.pop(0)['content'])
+
+        return memory
\ No newline at end of file