diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 72cbc40cd3..8df53a28c9 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -962,18 +962,10 @@ def list_gguf_variants(
             )
         )
 
-    # Sort: recommended first, then UD variants by size, then non-UD by size.
-    best_file = _pick_best_gguf([v.filename for v in variants])
-    best_quant = _extract_quant_label(best_file) if best_file else None
-
-    def _sort_key(v: GgufVariantInfo) -> tuple:
-        is_recommended = v.quant == best_quant
-        is_ud = v.quant.startswith("UD-")
-        # (0, ...) recommended | (1, 0, -size) other UD | (1, 1, -size) non-UD
-        # Negative size so largest (best quality) appears first.
-        return (0 if is_recommended else 1, 0 if is_ud else 1, -v.size_bytes)
-
-    variants.sort(key = _sort_key)
+    # Sort by size descending (largest = best quality first).
+    # Recommended pinning and OOM demotion are handled client-side
+    # where GPU VRAM info is available.
+    variants.sort(key = lambda v: -v.size_bytes)
 
     return variants, has_vision
 
diff --git a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
index 1ce7f40cca..cc399c39c5 100644
--- a/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
+++ b/studio/frontend/src/components/assistant-ui/model-selector/pickers.tsx
@@ -206,7 +206,24 @@ function GgufVariantExpander({
     );
   }
 
-  if (!variants || variants.length === 0) {
+  const sortedVariants = useMemo(() => {
+    if (!variants) return variants;
+    return [...variants].sort((a, b) => {
+      const aIsRec = a.quant === defaultVariant;
+      const bIsRec = b.quant === defaultVariant;
+      if (aIsRec !== bIsRec) return aIsRec ? -1 : 1;
+
+      const aGb = a.size_bytes / (1024 ** 3);
+      const bGb = b.size_bytes / (1024 ** 3);
+      const aOom = gpuGb != null && gpuGb > 0 && aGb > 0 && checkVramFit(aGb, gpuGb) === "exceeds";
+      const bOom = gpuGb != null && gpuGb > 0 && bGb > 0 && checkVramFit(bGb, gpuGb) === "exceeds";
+      if (aOom !== bOom) return aOom ? 1 : -1;
+
+      return b.size_bytes - a.size_bytes;
+    });
+  }, [variants, defaultVariant, gpuGb]);
+
+  if (!sortedVariants || sortedVariants.length === 0) {
     return (
       <div className="px-5 py-2 text-xs text-muted-foreground">
         No GGUF variants found.
@@ -224,7 +241,7 @@ function GgufVariantExpander({
           <span className="text-[9px] font-medium text-blue-400">Vision</span>
         )}
       </div>
-      {variants.map((v) => {
+      {sortedVariants.map((v) => {
         const sizeGb = v.size_bytes / (1024 ** 3);
         const fitStatus = gpuGb != null && gpuGb > 0 && sizeGb > 0
           ? checkVramFit(sizeGb, gpuGb)