diff --git a/lib/ai/vip98k-models.ts b/lib/ai/vip98k-models.ts
index 106cb3654..12337489b 100644
--- a/lib/ai/vip98k-models.ts
+++ b/lib/ai/vip98k-models.ts
@@ -113,7 +113,7 @@ const DEFAULT_ASPECT_RATIOS = [
 ] as const;
 
 const DEFAULT_VIDEO_RATIOS = ["16:9", "9:16", "1:1", "4:3", "3:4"] as const;
-const DEFAULT_IMAGE_SIZES = ["1K", "2K", "4K"] as const;
+const DEFAULT_IMAGE_SIZE = "1K";
 
 const TYPE_SORT_ORDER: Record<string, number> = {
   image: 100,
@@ -505,6 +505,13 @@ function normalizeImageSize(value: string): string | null {
   }
   const kMatch = normalized.match(/^(\d+(?:\.\d+)?)\s*K$/);
   if (kMatch?.[1]) return `${Number(kMatch[1]).toString()}K`;
+  const pixelMatch = normalized.match(/^(\d+)\s*[X*×]\s*(\d+)$/);
+  if (pixelMatch?.[1] && pixelMatch[2]) {
+    const longestEdge = Math.max(Number(pixelMatch[1]), Number(pixelMatch[2]));
+    if (longestEdge <= 1024) return "1K";
+    if (longestEdge <= 2048) return "2K";
+    if (longestEdge <= 4096) return "4K";
+  }
   return null;
 }
 
@@ -528,10 +535,17 @@ function getSupportedResolutions(model: Vip98kCatalogModel) {
 }
 
 function getSupportedImageSizes(model: Vip98kCatalogModel) {
-  const sizes = getSupportedResolutions(model)
+  const schemaSizes = getInvocationCandidates(model).flatMap((invocation) => {
+    const properties =
+      toRecord(toRecord(invocation.input_schema)?.properties) || {};
+    return ["size", "image_size", "imageSize", "resolution"].flatMap(
+      (key) => parseStringArray(toRecord(properties[key])?.enum),
+    );
+  });
+  const sizes = [...getSupportedResolutions(model), ...schemaSizes]
     .map(normalizeImageSize)
     .filter((value): value is string => Boolean(value));
-  return sizes.length ? sizes : [...DEFAULT_IMAGE_SIZES];
+  return sizes.length ? Array.from(new Set(sizes)) : [DEFAULT_IMAGE_SIZE];
 }
 
 function getSupportedVideoResolutions(model: Vip98kCatalogModel) {
diff --git a/tests/routes/vip98k-reference-routing.test.ts b/tests/routes/vip98k-reference-routing.test.ts
index 00a67610a..5c06fb4fc 100644
--- a/tests/routes/vip98k-reference-routing.test.ts
+++ b/tests/routes/vip98k-reference-routing.test.ts
@@ -479,7 +479,15 @@ test.describe("VIP98K reference-image routing", () => {
                 aspect_ratio: { type: "string" },
                 file_urls: { type: "array", maxItems: 8 },
                 resolution: { type: "string" },
-                size: { type: "string", enum: ["auto", "1024x1024"] },
+                size: {
+                  type: "string",
+                  enum: [
+                    "auto",
+                    "1024x1024",
+                    "2048x2048",
+                    "4096x4096",
+                  ],
+                },
               },
             },
             result_method: "GET",
@@ -589,6 +597,12 @@ test.describe("VIP98K reference-image routing", () => {
             n: 1,
             prompt: "Generate",
           },
+          input_schema: {
+            properties: {
+              image: { type: "array" },
+              size: { type: "string" },
+            },
+          },
         },
       },
     ]);
@@ -601,6 +615,14 @@ test.describe("VIP98K reference-image routing", () => {
       capabilities: {
         supportsMultiImageReference: true,
         maxReferenceImages: 8,
+        supportedImageSizes: ["1K"],
+      },
+      imageGenConfig: {
+        supportsSize: true,
+        supports1K: true,
+        supports2K: false,
+        supports4K: false,
+        supportedImageSizes: ["1K"],
       },
     });
   });