Studio: tag MLX loaded models as MLX instead of Base in chat (#6067)

* Studio: tag MLX loaded models as MLX instead of Base in chat

* Studio: tag MLX named hub defaults via name heuristic

studio/backend/core/inference/orchestrator.py

@@ -705,6 +705,7 @@ class InferenceOrchestrator:
                     self.models[self.active_model_name] = {
                         "is_vision": model_info.get("is_vision", False),
                         "is_lora": model_info.get("is_lora", False),
+                        "is_mlx": model_info.get("is_mlx", False),
                         "display_name": model_info.get("display_name", model_name),
                         "is_audio": model_info.get("is_audio", False),
                         "audio_type": model_info.get("audio_type"),

studio/backend/core/inference/worker.py

@@ -342,6 +342,8 @@ def _handle_load(backend, config: dict, resp_queue: Any) -> None:
                 "is_vision": mc.is_vision,
                 "is_lora": mc.is_lora,
                 "is_gguf": False,
+                # MLX backend sets device="mlx"; lets the UI tag MLX models.
+                "is_mlx": getattr(backend, "device", None) == "mlx",
                 "is_audio": getattr(mc, "is_audio", False),
                 "audio_type": getattr(mc, "audio_type", None),
                 "has_audio_input": getattr(mc, "has_audio_input", False),

studio/backend/models/models.py

@@ -76,6 +76,9 @@ class ModelDetails(BaseModel):
     is_gguf: bool = Field(
         False, description = "Whether model is a GGUF model (llama.cpp format)"
     )
+    is_mlx: bool = Field(
+        False, description = "Whether model is served via the MLX backend (Apple Silicon)"
+    )
     is_audio: bool = Field(False, description = "Whether model is a TTS audio model")
     audio_type: Optional[str] = Field(
         None, description = "Audio codec type: snac, csm, bicodec, dac"

studio/backend/routes/models.py

@@ -1446,6 +1446,15 @@ async def browse_folders(
     )
 
 
+def _looks_like_mlx_repo(model_id: str) -> bool:
+    """Name heuristic for unloaded models, mirrors the -GGUF suffix check.
+    Tokenized so MLX only matches as a whole name segment."""
+    if model_id.lower().startswith("mlx-community/"):
+        return True
+    tail = model_id.split("/")[-1]
+    return "MLX" in _re.split(r"[-_.]", tail.upper())
+
+
 @router.get("/list")
 async def list_models(
     current_subject: str = Depends(get_current_subject),
@@ -1471,6 +1480,7 @@ async def list_models(
                 name = model_name.split("/")[-1] if "/" in model_name else model_name,
                 is_vision = _is_vision,
                 is_lora = model_data.get("is_lora", False),
+                is_mlx = model_data.get("is_mlx", False),
                 is_audio = model_data.get("is_audio", False),
                 audio_type = _audio_type,
                 has_audio_input = model_data.get("has_audio_input", False),
@@ -1498,13 +1508,18 @@ async def list_models(
         all_models = []
         seen_ids = set()
 
+        # Prefer loaded entries for duplicate ids so runtime flags
+        # (is_mlx, is_vision, is_audio, ...) are not lost.
+        loaded_by_id = {model_info.id: model_info for model_info in loaded_models}
+
         # Add default models
         for model_id in default_models:
             if model_id not in seen_ids:
-                model_info = ModelDetails(
+                model_info = loaded_by_id.get(model_id) or ModelDetails(
                     id = model_id,
                     name = model_id.split("/")[-1] if "/" in model_id else model_id,
                     is_gguf = model_id.upper().endswith("-GGUF"),
+                    is_mlx = _looks_like_mlx_repo(model_id),
                 )
                 all_models.append(model_info)
                 seen_ids.add(model_id)

studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts

@@ -94,16 +94,25 @@ function describeModel(model: {
   is_lora?: boolean;
   is_vision?: boolean;
   is_gguf?: boolean;
+  is_mlx?: boolean;
   is_audio?: boolean;
   has_audio_input?: boolean;
 }): string | undefined {
   const tags: string[] = [];
   if (model.is_gguf) tags.push("GGUF");
+  if (model.is_mlx) tags.push("MLX");
   if (model.is_lora) tags.push("LoRA");
   if (model.is_vision) tags.push("Vision");
   if (model.is_audio) tags.push("Audio");
   if (model.has_audio_input) tags.push("Audio Input");
-  if (!model.is_lora && !model.is_vision && !model.is_gguf && !model.is_audio && !model.has_audio_input)
+  if (
+    !model.is_lora &&
+    !model.is_vision &&
+    !model.is_gguf &&
+    !model.is_mlx &&
+    !model.is_audio &&
+    !model.has_audio_input
+  )
     tags.push("Base");
   return tags.join(" · ");
 }
@@ -114,6 +123,7 @@ function toChatModelSummary(model: {
   is_lora?: boolean;
   is_vision?: boolean;
   is_gguf?: boolean;
+  is_mlx?: boolean;
   is_audio?: boolean;
   audio_type?: string | null;
   has_audio_input?: boolean;
@@ -125,6 +135,7 @@ function toChatModelSummary(model: {
     isLora: Boolean(model.is_lora),
     isVision: Boolean(model.is_vision),
     isGguf: Boolean(model.is_gguf),
+    isMlx: Boolean(model.is_mlx),
     isAudio: Boolean(model.is_audio),
     audioType: model.audio_type ?? null,
     hasAudioInput: Boolean(model.has_audio_input),

studio/frontend/src/features/chat/types/api.ts

@@ -7,6 +7,7 @@ export interface BackendModelDetails {
   is_vision?: boolean;
   is_lora?: boolean;
   is_gguf?: boolean;
+  is_mlx?: boolean;
   is_audio?: boolean;
   audio_type?: string | null;
   has_audio_input?: boolean;

studio/frontend/src/features/chat/types/runtime.ts

@@ -44,6 +44,7 @@ export interface ChatModelSummary {
   isVision: boolean;
   isLora: boolean;
   isGguf?: boolean;
+  isMlx?: boolean;
   isAudio?: boolean;
   audioType?: string | null;
   hasAudioInput?: boolean;