diff --git a/docs-site/content/cookbook/minimax-h3/open-h3-regenerate-2k.mdx b/docs-site/content/cookbook/minimax-h3/open-h3-regenerate-2k.mdx
new file mode 100644
index 000000000..ad9ab233b
--- /dev/null
+++ b/docs-site/content/cookbook/minimax-h3/open-h3-regenerate-2k.mdx
@@ -0,0 +1,7 @@
+---
+title: Open-H3-Regenerate-2K
+description: Configure Open-H3-Regenerate-2K serving, which regenerates a 768P MiniMax-H3 draft at 2560x1440.
+mode: frame
+---
+
+<Cookbook modelId="open-h3-regenerate-2k" />
diff --git a/docs-site/lib/cookbook-pages.ts b/docs-site/lib/cookbook-pages.ts
index 5c3d6b4ed..16b66193f 100644
--- a/docs-site/lib/cookbook-pages.ts
+++ b/docs-site/lib/cookbook-pages.ts
@@ -133,6 +133,7 @@ const backendModels: Record<string, string> = {
   "FastVideo/FastVideo-Minimax-FastH3-Preview-v0.2": "fasth3-preview",
   "FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree": "fasth3-4step",
   "FastVideo/FastH3-OmniRef-NVFP4-QAD": "fasth3-omniref",
+  "Efficient-Large-Model/Open-H3-Regenerate-2K": "open-h3-regenerate-2k",
   "FastVideo/FastWan2.1-T2V-1.3B-Diffusers": "fastwan21",
   "Wan-AI/Wan2.1-I2V-14B-480P-Diffusers": "wan21-i2v",
   "FastVideo/LTX2-Distilled-Diffusers": "ltx2-distilled",
diff --git a/docs-site/navigation.json b/docs-site/navigation.json
index db8f15868..f7d673d61 100644
--- a/docs-site/navigation.json
+++ b/docs-site/navigation.json
@@ -174,6 +174,10 @@
             "label": "FastH3 4-Step V1",
             "href": "/cookbook/minimax-h3/fasth3-4step"
           },
+          {
+            "label": "Open-H3-Regenerate-2K",
+            "href": "/cookbook/minimax-h3/open-h3-regenerate-2k"
+          },
           {
             "label": "Recipes and setup",
             "href": "/cookbook/minimax-h3"
diff --git a/docs/assets/cookbook-recipes.json b/docs/assets/cookbook-recipes.json
index 4eac9fe00..a7d537af3 100644
--- a/docs/assets/cookbook-recipes.json
+++ b/docs/assets/cookbook-recipes.json
@@ -764,7 +764,7 @@
       "id": "open-h3-regenerate-2k",
       "family": "minimax_h3",
       "stage": "inference",
-      "task": "Video to 2K video",
+      "task": "Video super-resolution (2K regeneration)",
       "label": "Open-H3-Regenerate-2K",
       "summary": "Regenerate a 768P MiniMax-H3 draft at 2560x1440 with NVlabs' three-step refiner. The output is silent.",
       "model": "Efficient-Large-Model/Open-H3-Regenerate-2K",