@@ -767,6 +767,17 @@ msgstr "Не найдено лицо на картинке. Попробуйте
 msgid "The input image may contain real person."
 msgstr "Загруженное изображение может содержать реального человека."
 
+#: ml_model/exceptions.py:185
+msgid "not specified"
+msgstr "не указана"
+
+#: ml_model/exceptions.py:187
+msgid ""
+"Version \"%(version)s\" is not available. Available versions: "
+"%(available_versions)s."
+msgstr ""
+"Версия «%(version)s» недоступна. Доступные версии: %(available_versions)s."
+
 #: ml_model/models.py:18 ml_model/models.py:38 ml_model/models.py:70
 #: ml_model/models.py:182 tools/media/models.py:67
 msgid "Slug"
@@ -33,6 +33,7 @@ from ml_model.exceptions import (
     CorruptedFileError,
     FileExtensionNotSupported,
     FileUploadUnsupported,
+    ModelVersionNotAvailable,
 )
 from ml_model.models import ModelConfiguration, NeuronModel
 from ml_model.services.EmbeddingService import EmbeddingService
@@ -104,7 +105,9 @@ class Chatgpt(SimpleService):
     ) -> list[Message]:
         start_time = time.time()
         info = input_message.info.copy()
-        model_name = info.pop('version', 'gpt-4o')
+        model_name = info.pop('version', None)
+        if model_name is None or model_name not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(model_name, self.TOKENS_COST)
         user_system_prompt = info.pop('system_prompt', '')
         input_content = [{'type': 'text', 'text': input_message.content or ''}]
         file = input_message.file
@@ -145,8 +148,6 @@ class Chatgpt(SimpleService):
                 'top_p': info.pop('top_p', 0.5),
             }
             self.llm.tiktoken_model_name = 'gpt-4'
-            if model_name not in self.TOKENS_COST.keys():
-                raise Exception(_('No matching version found'))
             chat_history = self.get_chat_history(model_name=model_name)
             chat_history.add_message(HumanMessage(content=input_message.content))
             conversation = RunnableWithMessageHistory(
@@ -14,6 +14,7 @@ from ml_model.exceptions import (
 )
 from ml_model.models import NeuronModel
 from ml_model.services import Chatgpt
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.EmbeddingService import EmbeddingService
 from ml_model.services.FileService import FileProcessingService
 from poller.models import Proxy
@@ -99,7 +100,9 @@ class Chatgpt_5(Chatgpt):
     ) -> list[Message]:
         start_time = time.time()
         info = input_message.info.copy()
-        model_name = info.pop('version', 'gpt-5')
+        model_name = info.pop('version', None)
+        if model_name is None or model_name not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(model_name, self.TOKENS_COST)
         user_system_prompt = info.pop('system_prompt', '')
         input_content = [{'type': 'text', 'text': input_message.content or ''}]
         file = input_message.file
@@ -22,6 +22,7 @@ from ml_model.exceptions import (
 )
 from ml_model.models import NeuronModel
 from ml_model.services import Chatgpt
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.EmbeddingService import EmbeddingService
 from ml_model.services.FileService import FileProcessingService
 from poller.models import Proxy
@@ -113,7 +114,9 @@ class Chatgpt_5_4(Chatgpt):
     ) -> list[Message]:
         start_time = time.time()
         info = input_message.info.copy()
-        model_name = info.pop('version', 'gpt-5.4')
+        model_name = info.pop('version', None)
+        if model_name is None or model_name not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(model_name, self.TOKENS_COST)
         user_system_prompt = info.pop('system_prompt', '')
         plan_info = self.store.user.payment_plan
         is_regular_user = self.store.user.account_type == 'regular'
@@ -155,7 +155,7 @@ class Chatgpt_5_5(Chatgpt):
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
         info = input_message.info.copy()
-        model_name = info.pop('version', 'gpt-5.5')
+        model_name = 'gpt-5.5'
         user_system_prompt = info.pop('system_prompt', '')
         file = input_message.file
         image = None
@@ -9,7 +9,7 @@ import filetype
 from PIL import Image
 
 from messages.models import Message
-from ml_model.exceptions import FileExtensionNotSupported
+from ml_model.exceptions import FileExtensionNotSupported, ModelVersionNotAvailable
 from ml_model.services.EmbeddingService import EmbeddingService
 from ml_model.services.FileService import FileProcessingService
 from ml_model.services.base import SimpleService
@@ -65,7 +65,10 @@ class Claude(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = f'anthropic/{input_message.info.pop("version", "claude-sonnet-4.6")}'
+        version_slug = input_message.info.pop('version', None)
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        version = f'anthropic/{version_slug}'
         system_prompt = input_message.info.pop('system_prompt', '')
         callback_data = {'provider': {'order': ['Anthropic']}, **input_message.info}
         messages = [
@@ -3,6 +3,7 @@ from _decimal import Decimal
 from datetime import timedelta
 
 from messages.models import Message
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import openrouter_run
 from tools.chats.models import Chat
@@ -54,7 +55,9 @@ class Deepseek(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         info = input_message.info.copy()
-        version = info.pop('version')
+        version = info.pop('version', None)
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         system_prompt = info.pop('system_prompt', '')
 
         callback_data = {
@@ -10,6 +10,7 @@ from messages.models import Message
 from ml_model.adapters.bytedance_model_ark import BytedanceContentType, BytedanceModelArkAdapter
 from ml_model.exceptions import FileExtensionNotSupported
 from ml_model.services.FileService import FileProcessingService
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import bytedance_model_ark_run
 from tools.chats.models import Chat
@@ -72,7 +73,9 @@ class Dola_Seed(SimpleService):
         return msgs
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'seed-2-0-pro')
+        version = input_message.info.pop('version', None)
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         callback_data = {
             "reasoning_effort": "minimal",
             **input_message.info,
@@ -73,7 +73,7 @@ class Flux(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = input_message.info.get('version')
+        version = 'flux-schnell'
         user_prompt = self.translate_prompt(input_message.content)
         callback_data = dict(
             {
@@ -84,7 +84,7 @@ class Flux(SimpleService):
             }
         )
         runner = replicate_run(
-            f'{self._CALLBACK_BASE}{callback_data.get("version", "flux-schnell")}',
+            f'{self._CALLBACK_BASE}{version}',
             callback_data,
         )
         images = runner if isinstance(runner, list) else [runner]
@@ -14,6 +14,7 @@ from django.core.files.images import get_image_dimensions
 from messages.models import Message
 from ml_model.exceptions import PredictionInterruptedError, RequestBlocked, GenerationException, \
     FileExtensionNotSupported
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -67,7 +68,9 @@ class Flux_2(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = input_message.info.get('version', 'flux-2-pro')
+        version = input_message.info.get('version')
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         width = input_message.info.pop('width', 1024)
         height = input_message.info.pop('height', 1024)
         output_mp = math.ceil((width*height) / 1_000_000)
@@ -70,7 +70,7 @@ class Fluxkrea(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = input_message.info.get('version')
+        version = 'flux-krea-dev'
         callback_data = dict(
             {
                 'prompt': self.translate_prompt(input_message.content),
@@ -85,7 +85,7 @@ class Fluxkrea(SimpleService):
             input_message.file.close()
             callback_data.update({'image': image})
         runner = replicate_run(
-            f'{self._CALLBACK_BASE}{callback_data.get("version", "flux-krea-dev")}',
+            f'{self._CALLBACK_BASE}{version}',
             callback_data,
         )
         images = runner if isinstance(runner, list) else [runner]
@@ -1,18 +1,17 @@
 import time
-import httpx
-
-from backend import settings
-from decimal import Decimal
 from datetime import timedelta
+from decimal import Decimal
 from io import BytesIO
 from typing import Any
 
+import httpx
 from django.core.files import File
 
+from backend import settings
 from messages.models import Message
-
-from ml_model.exceptions import ModelTimeoutError, GenerationException
+from ml_model.exceptions import GenerationException, ModelTimeoutError
 from ml_model.services.base import SimpleService
+from poller.models import Proxy
 
 
 class Fluxlorafast(SimpleService):
@@ -43,12 +42,27 @@ class Fluxlorafast(SimpleService):
     ) -> list[Message]:
         messages: list[Message] = []
         for image in images:
+            for proxy in Proxy.objects.all():
+                client = httpx.Client(
+                    base_url='https://queue.fal.run',
+                    headers={'Authorization': f'Key {settings.FAL_API_KEY}'},
+                    timeout=600,
+                    proxy=f'{proxy.protocol}://{proxy.address}',
+                )
+
+                try:
+                    file = File(BytesIO(client.get(image).content), '.png')
+                except Exception:
+                    continue
+
+                break
+
             messages.append(
                 Message(
                     content_object=self.store,
                     elapsed_time=time,
                     content=prompt,
-                    file=File(BytesIO(httpx.get(image).content), '.png'),
+                    file=file,
                 )
             )
         if save:
@@ -64,20 +78,19 @@ class Fluxlorafast(SimpleService):
             '4:3': 'landscape_4_3',
             '16:9': 'landscape_16_9',
         }
-        requests_number = 0
         start_time = time.time()
-        version = input_message.info.get('version')
+        version = 'flux-lora'
         translated_prompt = self.translate_prompt(input_message.content)
         callback_data = dict(
             {
                 'prompt': f'in style of raif3_corporate Isometric illustration, '
-                          f'contemporary vector art style, 3/4 perspective view: {translated_prompt}',
+                f'contemporary vector art style, 3/4 perspective view: {translated_prompt}',
                 'model_version': 'fb90c17a-d410-41e7-9961-dc7c687bc627',
                 'image_size': sizes.get(input_message.info.get('image_size', '1:1')),
                 'loras': [
                     {
                         'path': 'https://v3.fal.media/files/elephant/JthCZoCdAr7'
-                                'LqnOiNNVCC_pytorch_lora_weights.safetensors'
+                        'LqnOiNNVCC_pytorch_lora_weights.safetensors'
                     }
                 ],
                 'guidance_scale': 5,
@@ -85,29 +98,53 @@ class Fluxlorafast(SimpleService):
                 'num_images': input_message.info.get('num_images', 4),
             }
         )
-        client = httpx.Client(
-            base_url="https://queue.fal.run",
-            headers={"Authorization": f"Key {settings.FAL_API_KEY}"},
-            timeout=600,
-        )
-        result = client.post(
-            f'fal-ai/{version}',
-            json={'prompt': input_message.content, **callback_data},
-        ).json()
-        try:
-            while True:
-                status = client.get(result['status_url']).json()
-                if status.get('status') == 'COMPLETED':
-                    break
-                requests_number += 1
-                if requests_number == 271:
-                    raise ModelTimeoutError
-                time.sleep(1/3)
-        except Exception as exc:
-            raise GenerationException from exc
+
+        final_result = self._request(version, input_message, callback_data)
         process_time = timedelta(seconds=(time.time() - start_time))
-        final_result = client.get(result['response_url']).json()
+
         images = [img['url'] for img in final_result['images']]
         self.handle_invoice(input_message.content_object.model, input_message=input_message, version=version)
         msgs = self.save_results(input_message.content, images, process_time, save)
         return msgs
+
+    def _request(self, version: str, input_message: Message, callback_data: dict) -> dict:
+        proxies = Proxy.objects.all()
+        for proxy in proxies:
+            requests_number = 0
+            client = httpx.Client(
+                base_url='https://queue.fal.run',
+                headers={'Authorization': f'Key {settings.FAL_API_KEY}'},
+                timeout=600,
+                proxy=f'{proxy.protocol}://{proxy.address}',
+            )
+            result = client.post(
+                f'fal-ai/{version}',
+                json={'prompt': input_message.content, **callback_data},
+            ).json()
+
+            is_success = False
+
+            while True:
+                if requests_number == 271 / len(proxies):
+                    break
+
+                try:
+                    status = client.get(result['status_url']).json()
+                except Exception:
+                    requests_number += 1
+                    continue
+                if status.get('status') == 'COMPLETED':
+                    is_success = True
+                    break
+
+                time.sleep(1 / 3)
+
+            if is_success:
+                break
+
+        if not is_success:
+            raise GenerationException from ModelTimeoutError
+
+        final_result = client.get(result['response_url']).json()
+
+        return final_result
@@ -14,6 +14,7 @@ from ml_model.models import (
     ModelInput,
     NeuronModel,
 )
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -84,6 +85,8 @@ class Fluxproultra(SimpleService):
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
         version = input_message.info.get('version')
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         callback_data = dict(
             {
                 'prompt': self.translate_prompt(input_message.content),
@@ -9,7 +9,7 @@ from django.db.models.fields.files import FieldFile
 from PIL import Image
 
 from messages.models import Message
-from ml_model.exceptions import FileExtensionNotSupported
+from ml_model.exceptions import FileExtensionNotSupported, ModelVersionNotAvailable
 from ml_model.services.EmbeddingService import EmbeddingService
 from ml_model.services.FileService import FileProcessingService
 from ml_model.services.base import SimpleService
@@ -94,14 +94,17 @@ class Gemini(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         info = input_message.info.copy()
-        version = info.pop('version')
+        version_slug = info.pop('version', None)
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        version = f'google/{version_slug}'
         callback_data = {
             'provider': {'order': ['Google AI Studio']},
             **input_message.info,
         }
         messages = self.get_chat_history()
         messages.append({'role': 'user', 'content': input_message.content})
-        if version == 'google/gemini-3-flash-preview':
+        if version_slug == 'gemini-3-flash-preview':
             messages.insert(
                 0,
                 {
@@ -12,6 +12,7 @@ from messages.models import Message
 from ml_model.exceptions import CorruptedFileError, FileExtensionNotSupported
 from ml_model.services.EmbeddingService import EmbeddingService
 from ml_model.services.FileService import FileProcessingService
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import openrouter_run
 from poller.models import Proxy
@@ -66,7 +67,10 @@ class Gemini_3_1(SimpleService):
         return msgs
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.get('version', 'gemini-3.1-pro-preview:online')
+        version_slug = input_message.info.get('version')
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        model_slug = f'google/{version_slug}:online'
         callback_data = {
             'provider': {'order': ['Google AI Studio']},
             **input_message.info,
@@ -135,11 +139,11 @@ class Gemini_3_1(SimpleService):
             else:
                 raise FileExtensionNotSupported(supported_extensions)
         start_time = time.time()
-        result = openrouter_run(f'google/{version}:online', messages, callback_data, 'Gemini 3.1')
+        result = openrouter_run(model_slug, messages, callback_data, 'Gemini 3.1')
         process_time = timedelta(seconds=(time.time() - start_time))
         self.handle_invoice(
             input_message.content_object.model,
-            version=version,
+            version=version_slug,
             input_tokens=result[1],
             output_tokens=result[2],
             embedding_tokens=embedding_tokens,
@@ -44,7 +44,7 @@ class Glm_4_7(SimpleService):
         return msgs
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'glm-4-7-251222')
+        version = 'glm-4-7-251222'
         callback_data = {
             "reasoning_effort": "minimal",
             **input_message.info,
@@ -9,7 +9,7 @@ from pathlib import Path
 import filetype
 from PIL import Image
 from ml_model.services.FileService import FileProcessingService
-from ml_model.exceptions import FileExtensionNotSupported, CorruptedFileError
+from ml_model.exceptions import CorruptedFileError, FileExtensionNotSupported
 from ml_model.services.EmbeddingService import EmbeddingService
 from payments.exceptions.insufficient_balance import InsufficientBalance
 from payments.selectors.payment_plan_selector import PaymentPlanSelector
@@ -68,7 +68,7 @@ class Grok(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = f'x-ai/{input_message.info.pop("version", "grok-4.3")}'
+        version = 'x-ai/grok-4.3'
         callback_data = {**input_message.info}
         messages = self.get_chat_history()
         messages.append({'role': 'user', 'content': input_message.content})
@@ -11,6 +11,7 @@ from django.core.files import File
 
 from messages.models import Message
 from ml_model.exceptions import RequestBlocked, GenerationException
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -53,7 +54,9 @@ class Hailuo(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.get('version', 'hailuo-2.3-fast')
+        version = input_message.info.get('version')
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         resolution = input_message.info.get('resolution', '768p')
         if (
             (balance := PaymentPlanSelector(self.store.user).get_current_balance())
@@ -67,7 +67,7 @@ class Ideogram(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = input_message.info.get('version')
+        version = 'ideogram-v3-turbo'
         if (
             input_message.info.get('style_preset', 'None') != 'None'
             and input_message.info.get('style_type', 'None') not in ('None', 'Auto', 'General')
@@ -80,7 +80,7 @@ class Ideogram(SimpleService):
             }
         )
         runner = replicate_run(
-            f'{self._CALLBACK_BASE}{callback_data.get("version", "ideogram-v3-turbo")}',
+            f'{self._CALLBACK_BASE}{version}',
             callback_data,
         )
         images = runner if isinstance(runner, list) else [runner]
@@ -63,7 +63,7 @@ class Leonardo(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = input_message.info.get('version', 'lucid-origin')
+        version = 'lucid-origin'
         generation_mode = input_message.info.get('generation_mode', 'standard')
         num_images = input_message.info.get('num_images', 1)
         callback_data = dict(
@@ -73,7 +73,7 @@ class Leonardo(SimpleService):
             }
         )
         runner = replicate_run(
-            f'{self._CALLBACK_BASE}{callback_data.get("version", "lucid-origin")}',
+            f'{self._CALLBACK_BASE}{version}',
             callback_data,
         )
         images = runner if isinstance(runner, list) else [runner]
@@ -10,6 +10,7 @@ from datetime import timedelta
 from django.db.models.fields.files import FieldFile
 
 from messages.models import Message
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import openrouter_run
 from tools.chats.models import Chat
@@ -61,7 +62,10 @@ class Llama(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = f'meta-llama/{input_message.info.pop('version', 'llama-4-maverick')}'
+        version_slug = input_message.info.pop('version', None)
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        version = f'meta-llama/{version_slug}'
         callback_data = {'provider': {'order': ['DeepInfra']}, **input_message.info}
         messages = self.get_chat_history()
         messages.append({'role': 'user', 'content': input_message.content})
@@ -9,6 +9,7 @@ from django.core.files import File
 
 from messages.models import Message
 from ml_model.exceptions import RequestBlocked, GenerationException
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 from payments.exceptions.insufficient_balance import InsufficientBalance
@@ -37,7 +38,9 @@ class Lyria(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'lyria-3')
+        version = input_message.info.pop('version', None)
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         if (balance := PaymentPlanSelector(self.store.user).get_current_balance()) < (
             cost := self.TOKENS_COST[version]
         ):
@@ -42,7 +42,7 @@ class Minimaxvideo(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'video-01')
+        version = 'video-01'
         if (balance := PaymentPlanSelector(self.store.user).get_current_balance()) < self.TOKENS_COST[version]:
             raise InsufficientBalance(balance, self.TOKENS_COST[version])
         callback_data = dict({'prompt': input_message.content, **input_message.info})
@@ -54,8 +54,7 @@ class Mistral(SimpleService):
         return msgs
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        info = input_message.info.copy()
-        version = f'mistralai/{info.pop("version")}'
+        version = 'mistralai/mistral-small-3.1-24b-instruct'
         callback_data = {'provider': {'order': ['Parasail']}, **input_message.info}
         messages = self.get_chat_history()
         messages.append({'role': 'user', 'content': input_message.content})
@@ -16,6 +16,7 @@ from ml_model.exceptions import (
     ModelCouldNotInterpretPrompt,
     RequestBlocked,
 )
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -62,7 +63,9 @@ class Nanobanana(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = input_message.info.get('version', 'nano-banana')
+        version = input_message.info.get('version')
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         resolution = input_message.info.get('resolution', '2K') if version == 'nano-banana-pro' else None
         callback_data = dict(
             {
@@ -8,6 +8,7 @@ import httpx
 from django.conf import settings
 
 from messages.models import Message
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from poller.models import Proxy
 from tools.chats.models import Chat
@@ -64,7 +65,10 @@ class Perplexity(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = f'perplexity/{input_message.info.pop("version", "sonar")}'
+        version_slug = input_message.info.pop('version', None)
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        version = f'perplexity/{version_slug}'
         callback_data = {'provider': {'order': ['Perplexity']}, **input_message.info}
         messages = self.get_chat_history()
         messages.append({'role': 'user', 'content': input_message.content})
@@ -11,6 +11,7 @@ from django.core.files import File
 
 from messages.models import Message
 from ml_model.exceptions import GenerationException, RequestBlocked
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 from payments.exceptions.insufficient_balance import InsufficientBalance
@@ -42,7 +43,9 @@ class Prunaai(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.get('version', 'p-image')
+        version = input_message.info.get('version')
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         aspect_ratio = input_message.info.pop('aspect_ratio', 'custom')
         if input_message.file:
             version = 'p-image-edit'
@@ -4,6 +4,7 @@ from decimal import Decimal
 from typing import Any, Dict, Iterator
 
 from messages.models import Message
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import openrouter_run
 from tools.chats.models import Chat
@@ -49,7 +50,10 @@ class Qwen(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = f'qwen/{input_message.info.pop("version", "qwq-32b")}'
+        version_slug = input_message.info.pop('version', None)
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        version = f'qwen/{version_slug}'
         callback_data = {'provider': {'order': ['DeepInfra']}, **input_message.info}
         messages = self.get_chat_history()
         messages.insert(
@@ -4,6 +4,7 @@ from decimal import Decimal
 from typing import Any, Dict, Iterator
 
 from messages.models import Message
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import openrouter_run
 from tools.chats.models import Chat
@@ -53,7 +54,10 @@ class Qwen_235B(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = f'qwen/{input_message.info.pop("version", "qwen3-235b-a22b")}'
+        version_slug = input_message.info.pop('version', None)
+        if version_slug is None or version_slug not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version_slug, self.TOKENS_COST)
+        version = f'qwen/{version_slug}'
         callback_data = {'provider': {'order': ['DeepInfra']}, **input_message.info}
         messages = self.get_chat_history()
         messages.insert(
@@ -11,6 +11,7 @@ from messages.models import Message
 from ml_model.exceptions import FileExtensionNotSupported
 from ml_model.services.EmbeddingService import EmbeddingService
 from ml_model.services.FileService import FileProcessingService
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import openrouter_run
 from poller.models import Proxy
@@ -62,7 +63,10 @@ class Qwen_3_6(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
-        version = self.MODELS_LINKS[input_message.info.get('version', 'qwen3.6-flash')]
+        version_slug = input_message.info.get('version')
+        if version_slug is None or version_slug not in self.MODELS_LINKS:
+            raise ModelVersionNotAvailable(version_slug, self.MODELS_LINKS)
+        api_slug = self.MODELS_LINKS[version_slug]
         callback_data = {'max_tokens': self.MAX_OUTPUT_TOKENS, **input_message.info}
         messages = self.get_chat_history()
         messages.append({'role': 'user', 'content': input_message.content})
@@ -112,12 +116,12 @@ class Qwen_3_6(SimpleService):
                 ]
             else:
                 raise FileExtensionNotSupported(['PDF', 'DOC', 'DOCX', 'XLSX', 'JPG', 'JPEG', 'PNG', 'WEBP'])
-        model_slug = f'qwen/{version}:online'
+        model_slug = f'qwen/{api_slug}:online'
         result = openrouter_run(model_slug, messages, callback_data, 'Qwen 3.6')
         process_time = timedelta(seconds=(time.time() - start_time))
         self.handle_invoice(
             input_message.content_object.model,
-            version=version,
+            version=api_slug,
             input_tokens=result[1],
             output_tokens=result[2],
             embedding_tokens=embedding_tokens,
@@ -11,6 +11,7 @@ from django.core.files import File
 
 from messages.models import Message
 from ml_model.exceptions import RequestBlocked, GenerationException
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -44,7 +45,9 @@ class Ray(SimpleService):
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         duration = input_message.info.get('duration', 5)
-        version = input_message.info.get('version', 'ray-2-720p')
+        version = input_message.info.get('version')
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         if (balance := PaymentPlanSelector(self.store.user).get_current_balance()) < (
             cost := self.TOKENS_COST[version] * duration
         ):
@@ -14,6 +14,7 @@ from ml_model.models import (
     ModelParameter,
     ModelVersion,
 )
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -178,14 +179,15 @@ class Recraft(SimpleService):
             'линогравюра': 'linocut',
         }
         start_time = time.time()
-        extension = (
-            '.svg' if input_message.info.get('version', 'recraft-v3') == self.versions[1].slug else '.png'
-        )
+        version = input_message.info.get('version')
+        if version is None or version not in self.payment_rules:
+            raise ModelVersionNotAvailable(version, self.payment_rules)
+        extension = '.svg' if version == self.versions[1].slug else '.png'
         size = self._get_size(
             input_message.info.pop('width', 1024),
             input_message.info.pop('height', 1024),
         )
-        callback_url = f'recraft-ai/{input_message.info.get("version", "recraft-v3")}'
+        callback_url = f'recraft-ai/{version}'
         callback_data = dict(
             {
                 'prompt': (
@@ -47,7 +47,7 @@ class Runway(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'gen4-turbo')
+        version = 'gen4-turbo'
         duration = input_message.info.get('duration', 5)
         file = input_message.file
         if (
@@ -11,6 +11,7 @@ from django.core.files import File
 from messages.models import Message
 from ml_model.exceptions import RequestBlocked, GenerationException, FileExtensionNotSupported
 from ml_model.services.FileService import FileProcessingService
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -63,7 +64,9 @@ class Seedance(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'seedance-2.0')
+        version = input_message.info.pop('version', None)
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         resolution = input_message.info.get('resolution', '720p')
         duration = input_message.info.get('duration', 5)
         file = input_message.file or None
@@ -16,6 +16,7 @@ from messages.models import Message
 from ml_model.adapters.bytedance_model_ark import BytedanceContentType
 from ml_model.exceptions import FileExtensionNotSupported, InvalidParameterError
 from ml_model.services.FileService import FileProcessingService
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import bytedance_model_ark_run
 
@@ -100,7 +101,9 @@ class Seedance_2_Dreamina(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'dreamina-seedance-2-0')
+        version = input_message.info.pop('version', None)
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         resolution = input_message.info.get('resolution', '720p')
         duration = input_message.info.get('duration', 5)
         ratio = input_message.info.get('ratio', '16:9')
@@ -12,6 +12,7 @@ from replicate.exceptions import ModelError
 from messages.models import Message
 from ml_model.adapters.bytedance_model_ark import BytedanceContentType
 from ml_model.exceptions import InvalidParameterError
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import bytedance_model_ark_run
 from payments.exceptions.insufficient_balance import InsufficientBalance
@@ -80,7 +81,9 @@ class Seedream(SimpleService):
         )):
             raise InsufficientBalance(balance, predicted)
 
-        version = input_message.info.pop('version', 'seedream-boosted')
+        version = input_message.info.pop('version', None)
+        if version is None or version not in self.TOKEN_COST:
+            raise ModelVersionNotAvailable(version, self.TOKEN_COST)
         size = input_message.info.get('size', '2K')
 
         if version == 'seedream-4.5' and size == '3K':
@@ -13,6 +13,7 @@ from django.core.files import File
 
 from messages.models import Message
 from ml_model.exceptions import ModelTimeoutError, RequestBlocked, UnsupportedSize
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from poller.models import Proxy
 
@@ -50,7 +51,9 @@ class Sora(SimpleService):
     def make(self, input_message: "Message", save: bool = True) -> list["Message"]:
         if input_message.content:
             for proxy in Proxy.objects.all():
-                version = input_message.info.pop('version', 'sora-2')
+                version = input_message.info.pop('version', None)
+                if version is None or version not in self.TOKENS_COST:
+                    raise ModelVersionNotAvailable(version, self.TOKENS_COST)
                 seconds = int(input_message.info.pop('seconds', 4))
                 callback_data = {
                     'prompt': input_message.content,
@@ -11,7 +11,7 @@ from django.conf import settings
 from django.core.files import File
 
 from messages.models import Message
-from ml_model.exceptions import RequestBlocked
+from ml_model.exceptions import ModelVersionNotAvailable, RequestBlocked
 from ml_model.services.base import SimpleService
 from poller.models import Proxy
 
@@ -76,7 +76,10 @@ class Stablediffusion(SimpleService):
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
         start_time = time.time()
         info = input_message.info.copy()
-        model_name = self.MODELS_LINKS[info.get('version', 'sd3')]
+        version = info.get('version')
+        if version is None or version not in self.MODELS:
+            raise ModelVersionNotAvailable(version, self.MODELS)
+        model_name = self.MODELS_LINKS[version]
         translated_prompt = self.translate_prompt(input_message.content)
         callback_data = {
             'prompt': translated_prompt,
@@ -11,6 +11,7 @@ from django.core.files import File
 
 from messages.models import Message
 from ml_model.exceptions import RequestBlocked, GenerationException
+from ml_model.exceptions import ModelVersionNotAvailable
 from ml_model.services.base import SimpleService
 from ml_model.tasks import replicate_run
 
@@ -41,7 +42,9 @@ class Veo(SimpleService):
         return [msg]
 
     def make(self, input_message: Message, save: bool = True) -> list[Message]:
-        version = input_message.info.pop('version', 'veo-3-fast')
+        version = input_message.info.pop('version', None)
+        if version is None or version not in self.TOKENS_COST:
+            raise ModelVersionNotAvailable(version, self.TOKENS_COST)
         if (balance := PaymentPlanSelector(self.store.user).get_current_balance()) < self.TOKENS_COST[
             version
         ]:
@@ -174,3 +174,18 @@ class FaceNotFoundError(Exception):
 class RealPersonDetectedError(Exception):
     def __str__(self) -> str:
         return _('The input image may contain real person.')
+
+
+class ModelVersionNotAvailable(Exception):
+    def __init__(self, version: str | None, available_versions: Iterable[str]) -> None:
+        self.version = version
+        self.available_versions = list(available_versions)
+
+    def __str__(self) -> str:
+        version_label = self.version if self.version is not None else _('not specified')
+        return _(
+            'Version "%(version)s" is not available. Available versions: %(available_versions)s.'
+        ) % {
+            'version': version_label,
+            'available_versions': ', '.join(self.available_versions),
+        }
@@ -27,13 +27,14 @@ from ml_model.exceptions import (
     FileTooLargeError,
     FileUploadUnsupported,
     ImageAnalysisError,
-    UnrecognizedFileError,
+    InvalidParameterError,
+    ModelVersionNotAvailable,
     PaidPlanRequiredError,
     PromptLengthExceeded,
     RequestBlocked,
     TemplateNotFound,
     TemplateUnknownException,
-    InvalidParameterError
+    UnrecognizedFileError,
 )
 from ml_model.services.base import SimpleService
 from payments.exceptions.insufficient_balance import InsufficientBalance
@@ -184,6 +185,7 @@ class MessagesAPIView(APIView):
                 FileUploadUnsupported,
                 UnrecognizedFileError,
                 InvalidParameterError,
+                ModelVersionNotAvailable,
             ) as exc:
                 return Response({'detail': f'{exc}'}, status=HTTP_400_BAD_REQUEST)
             except TemplateNotFound as exc:
@@ -17,17 +17,18 @@ from ml_model.exceptions import (
     FileExtensionNotSupported,
     FileNotProvided,
     FileTooLargeError,
-    UnrecognizedFileError,
     ImageAnalysisError,
     ImageContentNotFound,
     InvalidParameterError,
-    ModelCouldNotInterpretPrompt,
     InvalidStyleCombinationError,
+    ModelCouldNotInterpretPrompt,
+    ModelVersionNotAvailable,
     PromptLengthExceeded,
+    RealPersonDetectedError,
     RequestBlocked,
     ServiceHighDemandError,
+    UnrecognizedFileError,
     UnsupportedSize,
-    RealPersonDetectedError
 )
 from ml_model.models import NeuronModel
 from ml_model.services.base import SimpleService
@@ -190,6 +191,7 @@ class MediaAPIView(APIView):
                         ModelCouldNotInterpretPrompt,
                         InvalidStyleCombinationError,
                         InvalidParameterError,
+                        ModelVersionNotAvailable,
                         PromptLengthExceeded,
                         ExceededContextLengthError,
                         FileExtensionNotSupported,