diff --git a/ai_gateway/api/v2/code/completions.py b/ai_gateway/api/v2/code/completions.py
index d6a0b266cba74638f6e703dd03c93efbefd00e9d..b01bf5966646742d926718b9d5c93e89c1a793d1 100644
--- a/ai_gateway/api/v2/code/completions.py
+++ b/ai_gateway/api/v2/code/completions.py
@@ -13,6 +13,7 @@ from ai_gateway.api.snowplow_context import get_snowplow_code_suggestion_context
 from ai_gateway.api.v2.code.typing import (
     CompletionsRequestV1,
     CompletionsRequestV2,
+    CompletionsRequestV3,
     GenerationsRequestV1,
     GenerationsRequestV2,
     GenerationsRequestV3,
@@ -73,7 +74,7 @@ request_log = get_request_logger("codesuggestions")
 router = APIRouter()
 
 CompletionsRequestWithVersion = Annotated[
-    Union[CompletionsRequestV1, CompletionsRequestV2],
+    Union[CompletionsRequestV1, CompletionsRequestV2, CompletionsRequestV3],
     Body(discriminator="prompt_version"),
 ]
 
@@ -146,7 +147,7 @@ async def completions(
     except Exception as e:
         log_exception(e)
 
-    request_log.debug(
+    request_log.info(
         "code completion input:",
         model_name=payload.model_name,
         model_provider=payload.model_provider,
@@ -158,11 +159,14 @@ async def completions(
     )
 
     kwargs = {}
+
     if payload.model_provider == KindModelProvider.ANTHROPIC:
-        code_completions = completions_anthropic_factory()
+        code_completions = completions_anthropic_factory(
+            model__name=payload.model_name,
+        )
 
-        # We support the prompt version 2 only with the Anthropic models
-        if payload.prompt_version == 2:
+        # We support the prompt version 3 only with the Anthropic models
+        if payload.prompt_version == 3:
             kwargs.update({"raw_prompt": payload.prompt})
     elif payload.model_provider in (
         KindModelProvider.LITELLM,
diff --git a/ai_gateway/api/v2/code/typing.py b/ai_gateway/api/v2/code/typing.py
index 14fbeb1d4108fe56acee913e3360dd8e51f7ef22..f811d9fad902e8da1193a3fc9cd3c354891b2640 100644
--- a/ai_gateway/api/v2/code/typing.py
+++ b/ai_gateway/api/v2/code/typing.py
@@ -24,6 +24,7 @@ __all__ = [
     "CompletionsRequestV1",
     "GenerationsRequestV1",
     "CompletionsRequestV2",
+    "CompletionsRequestV3",
     "GenerationsRequestV2",
     "SuggestionsResponse",
     "StreamSuggestionsResponse",
@@ -104,6 +105,11 @@ class CompletionsRequestV2(CompletionsRequest):
     prompt: Optional[str] = None
 
 
+class CompletionsRequestV3(CompletionsRequest):
+    prompt_version: Literal[3]
+    prompt: Optional[list[Message]] = None
+
+
 class GenerationsRequestV2(GenerationsRequest):
     prompt_version: Literal[2]
     prompt: str
diff --git a/ai_gateway/api/v3/code/completions.py b/ai_gateway/api/v3/code/completions.py
index 858b8115c5216e258e03d1b1f15936a45879fd19..6fc5c19ea57c152fb11136786e11132d5626dd45 100644
--- a/ai_gateway/api/v3/code/completions.py
+++ b/ai_gateway/api/v3/code/completions.py
@@ -116,12 +116,15 @@ async def code_completion(
     code_context: list[CodeContextPayload] = None,
     snowplow_event_context: Optional[SnowplowEventContext] = None,
 ):
+    kwargs = {}
+
     if payload.model_provider == ModelProvider.ANTHROPIC:
-        engine = completions_anthropic_factory()
+        # TODO: As we migrate to v3 we can rewrite this to use prompt registry
+        engine = completions_anthropic_factory(model__name=payload.model_name)
+        kwargs.update({"raw_prompt": payload.prompt})
     else:
         engine = completions_legacy_factory()
 
-    kwargs = {}
     if payload.choices_count > 0:
         kwargs.update({"candidate_count": payload.choices_count})
 
diff --git a/ai_gateway/api/v3/code/typing.py b/ai_gateway/api/v3/code/typing.py
index c195557901793fd44bf119693aafae656a87584d..14cf8a46af8a4e1aeb0ae6a27ef4270e8ce37592 100644
--- a/ai_gateway/api/v3/code/typing.py
+++ b/ai_gateway/api/v3/code/typing.py
@@ -6,6 +6,7 @@ from pydantic import BaseModel, ConfigDict, Field, StringConstraints
 from starlette.responses import StreamingResponse
 
 from ai_gateway.code_suggestions import ModelProvider
+from ai_gateway.models import Message
 
 __all__ = [
     "CodeEditorComponents",
@@ -46,6 +47,8 @@ class EditorContentPayload(BaseModel):
 
 class EditorContentCompletionPayload(EditorContentPayload):
     choices_count: Optional[int] = 0
+    model_name: Optional[str] = None
+    prompt: Optional[str | list[Message]] = None
 
 
 class EditorContentGenerationPayload(EditorContentPayload):
diff --git a/ai_gateway/code_suggestions/base.py b/ai_gateway/code_suggestions/base.py
index 372aa5899f0098ced6b6d1aa72a960abf23baca8..1fb351f991dd021540aeb35513dd8d26a6ad74dc 100644
--- a/ai_gateway/code_suggestions/base.py
+++ b/ai_gateway/code_suggestions/base.py
@@ -51,6 +51,7 @@ USE_CASES_MODELS_MAP = {
     KindUseCase.CODE_COMPLETIONS: {
         KindAnthropicModel.CLAUDE_INSTANT_1_1,
         KindAnthropicModel.CLAUDE_INSTANT_1_2,
+        KindAnthropicModel.CLAUDE_3_5_SONNET,
         KindVertexTextModel.CODE_GECKO_002,
         KindVertexTextModel.CODESTRAL_2405,
         KindLiteLlmModel.CODEGEMMA,
diff --git a/ai_gateway/code_suggestions/completions.py b/ai_gateway/code_suggestions/completions.py
index ff216e31a2d4e40d24d07eabb9331ed7902649e4..04469e7772d6b5cae0fdb56dd787f3c9900add90 100644
--- a/ai_gateway/code_suggestions/completions.py
+++ b/ai_gateway/code_suggestions/completions.py
@@ -24,7 +24,7 @@ from ai_gateway.instrumentators import (
     TextGenModelInstrumentator,
     benchmark,
 )
-from ai_gateway.models import ModelAPICallError, ModelAPIError
+from ai_gateway.models import ChatModelBase, Message, ModelAPICallError, ModelAPIError
 from ai_gateway.models.agent_model import AgentModel
 from ai_gateway.models.base import TokensConsumptionMetadata
 from ai_gateway.models.base_text import (
@@ -143,7 +143,7 @@ class CodeCompletions:
         self,
         prefix: str,
         suffix: str,
-        raw_prompt: Optional[str] = None,
+        raw_prompt: Optional[str | list[Message]] = None,
         code_context: Optional[list] = None,
         context_max_percent: Optional[float] = None,
     ) -> Prompt:
@@ -168,7 +168,7 @@ class CodeCompletions:
         suffix: str,
         file_name: str,
         editor_lang: Optional[str] = None,
-        raw_prompt: Optional[str] = None,
+        raw_prompt: Optional[str | list[Message]] = None,
         code_context: Optional[list] = None,
         stream: bool = False,
         snowplow_event_context: Optional[SnowplowEventContext] = None,
@@ -196,6 +196,10 @@ class CodeCompletions:
                     params = {"prefix": prompt.prefix, "suffix": prompt.suffix}
 
                     res = await self.model.generate(params, stream)
+                elif isinstance(self.model, ChatModelBase):
+                    res = await self.model.generate(
+                        prompt.prefix, stream=stream, **kwargs
+                    )
                 else:
                     res = await self.model.generate(
                         prompt.prefix, prompt.suffix, stream, **kwargs
diff --git a/ai_gateway/code_suggestions/container.py b/ai_gateway/code_suggestions/container.py
index 4d52a57fdeafbbe5ff2675e306b66b38d0f5b263..d66bb54d6a2e3d319322dc8f37131ee4a0942e98 100644
--- a/ai_gateway/code_suggestions/container.py
+++ b/ai_gateway/code_suggestions/container.py
@@ -101,6 +101,7 @@ class ContainerCodeCompletions(containers.DeclarativeContainer):
     tokenizer = providers.Dependency(instance_of=PreTrainedTokenizerFast)
     vertex_code_gecko = providers.Dependency(instance_of=TextGenModelBase)
     anthropic_claude = providers.Dependency(instance_of=TextGenModelBase)
+    anthropic_claude_chat = providers.Dependency(instance_of=ChatModelBase)
     litellm = providers.Dependency(instance_of=TextGenModelBase)
     agent_model = providers.Dependency(instance_of=TextGenModelBase)
     snowplow_instrumentator = providers.Dependency(instance_of=SnowplowInstrumentator)
@@ -128,12 +129,7 @@ class ContainerCodeCompletions(containers.DeclarativeContainer):
 
     anthropic = providers.Factory(
         CodeCompletions,
-        model=providers.Factory(
-            anthropic_claude,
-            name=KindAnthropicModel.CLAUDE_INSTANT_1_2,
-            stop_sequences=["</new_code>", anthropic.HUMAN_PROMPT],
-            max_tokens_to_sample=128,
-        ),
+        model=providers.Factory(anthropic_claude_chat),
         tokenization_strategy=providers.Factory(
             TokenizerTokenStrategy, tokenizer=tokenizer
         ),
@@ -201,6 +197,7 @@ class ContainerCodeSuggestions(containers.DeclarativeContainer):
         tokenizer=tokenizer,
         vertex_code_gecko=models.vertex_code_gecko,
         anthropic_claude=models.anthropic_claude,
+        anthropic_claude_chat=models.anthropic_claude_chat,
         litellm=models.litellm,
         agent_model=models.agent_model,
         config=config,
diff --git a/tests/api/v2/test_v2_code.py b/tests/api/v2/test_v2_code.py
index c2c78f8baaa399f912c88ea5e0dbd09b008ca907..4b090e669eb84769e675a6ef1a9b3efbcc9cfeef 100644
--- a/tests/api/v2/test_v2_code.py
+++ b/tests/api/v2/test_v2_code.py
@@ -295,17 +295,17 @@ class TestCodeCompletions:
                 },
                 False,
             ),
-            # prompt version 2
+            # prompt version 3
             (
-                2,
+                3,
                 "anthropic",
-                "claude-instant-1.2",
+                "claude-3-5-sonnet-20240620",
                 "def search",
                 {
                     "id": "id",
                     "model": {
                         "engine": "anthropic",
-                        "name": "claude-instant-1.2",
+                        "name": "claude-3-5-sonnet-20240620",
                         "lang": "python",
                         "tokens_consumption_metadata": None,
                     },
@@ -406,10 +406,23 @@ class TestCodeCompletions:
                 }
             )
 
-            if mock_suggestions_engine == "anthropic":
-                code_completions_kwargs.update(
-                    {"raw_prompt": current_file["content_above_cursor"]}
-                )
+        if prompt_version == 3 and mock_suggestions_engine == "anthropic":
+            raw_prompt = [
+                {
+                    "role": "system",
+                    "content": "You are a code completion tool that performs Fill-in-the-middle. ",
+                },
+                {
+                    "role": "user",
+                    "content": "<SUFFIX>\n// write a function to find the max\n</SUFFIX>\n<PREFIX>\n\n\treturn min\n}\n</PREFIX>')]",
+                },
+            ]
+            data.update({"prompt": raw_prompt})
+            raw_prompt = [
+                Message(role=prompt["role"], content=prompt["content"])
+                for prompt in raw_prompt
+            ]
+            code_completions_kwargs.update({"raw_prompt": raw_prompt})
 
         response = mock_client.post(
             "/completions",
@@ -457,6 +470,7 @@ class TestCodeCompletions:
             "project_id": 278964,
             "model_provider": "anthropic",
             "current_file": current_file,
+            "model_name": "claude-3-5-sonnet-20240620",
         }
 
         code_completions_kwargs = {}
@@ -795,6 +809,7 @@ class TestCodeCompletions:
             "current_file": current_file,
             "stream": True,
             "context": context,
+            "model_name": "claude-3-5-sonnet-20240620",
         }
         if model:
             data["model_name"] = model
@@ -994,6 +1009,7 @@ class TestCodeCompletions:
             "project_id": 278964,
             "model_provider": "anthropic",
             "current_file": current_file,
+            "model_name": "claude-3-5-sonnet-20240620",
         }
 
         response = mock_client.post(
diff --git a/tests/api/v3/test_v3_code.py b/tests/api/v3/test_v3_code.py
index ceef3c24e4e5bbe3f051c9ac9eebd6a59d7029cb..cc5b5a357545ea45407027b3c39a88afd0865f79 100644
--- a/tests/api/v3/test_v3_code.py
+++ b/tests/api/v3/test_v3_code.py
@@ -173,7 +173,7 @@ class TestEditorContentCompletion:
                 },
                 {
                     "engine": "anthropic",
-                    "name": "claude-instant-1.2",
+                    "name": "claude-3-5-sonnet-20240620",
                     "lang": "python",
                 },
             ),
@@ -208,7 +208,7 @@ class TestEditorContentCompletion:
     def test_model_provider(
         self,
         mock_client: TestClient,
-        mock_anthropic: Mock,
+        mock_anthropic_chat: Mock,
         mock_code_gecko: Mock,
         model_provider: str,
         expected_code: int,
@@ -221,6 +221,17 @@ class TestEditorContentCompletion:
             "content_below_cursor": "\n",
             "language_identifier": "python",
             "model_provider": model_provider or None,
+            "model_name": "claude-3-5-sonnet-20240620",
+            "prompt": [
+                {
+                    "role": "system",
+                    "content": "You are a code completion tool that performs Fill-in-the-middle",
+                },
+                {
+                    "role": "user",
+                    "content": "<SUFFIX>\n// a function to find the max\n for \n</SUFFIX>\n<PREFIX>\n\n\treturn min\n}\n</PREFIX>",
+                },
+            ],
         }
 
         prompt_component = {
@@ -255,7 +266,7 @@ class TestEditorContentCompletion:
 
         assert body["metadata"]["model"] == expected_model
 
-        mock = mock_anthropic if model_provider == "anthropic" else mock_code_gecko
+        mock = mock_anthropic_chat if model_provider == "anthropic" else mock_code_gecko
 
         mock.assert_called
 
@@ -272,6 +283,7 @@ class TestEditorContentCompletion:
             "language_identifier": "python",
             "model_provider": "anthropic",
             "stream": True,
+            "model_name": "claude-3-5-sonnet-20240620",
         }
 
         prompt_component = {
@@ -319,6 +331,7 @@ class TestEditorContentCompletion:
             stream=True,
             code_context=None,
             snowplow_event_context=expected_snowplow_event,
+            raw_prompt=None,
         )
 
 
@@ -857,6 +870,7 @@ class TestIncomingRequest:
                                 "content_below_cursor": "",
                                 # FIXME: Forcing anthropic as vertex-ai is not working
                                 "model_provider": "anthropic",
+                                "model_name": "claude-3-5-sonnet-20240620",
                             },
                         },
                     ],
@@ -873,6 +887,7 @@ class TestIncomingRequest:
                                 "file_name": "test",
                                 "content_above_cursor": "def hello_world():",
                                 "content_below_cursor": "",
+                                "model_name": "claude-3-5-sonnet-20240620",
                             },
                         },
                     ],
@@ -889,6 +904,7 @@ class TestIncomingRequest:
                                 "file_name": "test",
                                 "content_above_cursor": "def hello_world():",
                                 "content_below_cursor": "",
+                                "model_name": "claude-3-5-sonnet-20240620",
                             },
                         },
                     ]
@@ -905,6 +921,7 @@ class TestIncomingRequest:
                             "payload": {
                                 "content_above_cursor": "def hello_world():",
                                 "content_below_cursor": "",
+                                "model_name": "claude-3-5-sonnet-20240620",
                             },
                         },
                     ],
diff --git a/tests/code_suggestions/test_container.py b/tests/code_suggestions/test_container.py
index b7ee620110637b5f184ed44d67a626f9b4dfee12..2cb7d7ba74f45f8316b66999006398650e697b5c 100644
--- a/tests/code_suggestions/test_container.py
+++ b/tests/code_suggestions/test_container.py
@@ -15,7 +15,10 @@ def test_container(mock_container: containers.DeclarativeContainer):
     generations = code_suggestions.generations
 
     assert isinstance(completions.vertex_legacy(), CodeCompletionsLegacy)
-    assert isinstance(completions.anthropic(), CodeCompletions)
+    assert isinstance(
+        completions.anthropic(model__name=KindAnthropicModel.CLAUDE_3_5_SONNET),
+        CodeCompletions,
+    )
     assert isinstance(
         completions.litellm_factory(model__name=KindLiteLlmModel.CODEGEMMA),
         CodeCompletions,