joshistoast · Oct 11, 2026 · Oct 11, 2026 · Oct 11, 2026 · Oct 11, 2026
diff --git a/docs/src/content/docs/Users Guide/03.Models/01.introduction.mdx b/docs/src/content/docs/Users Guide/03.Models/01.introduction.mdx
@@ -98,6 +98,18 @@ information on how to generate these keys. Once a valid API key has
 been provided, the list of models hosted by that service will appear
 as selections in the Image Generation panel.
 
+Opening an external model in the Model Manager shows an **External
+Provider** section where you can change the provider (one of the
+registered providers) and provider model ID it routes to, the modes
+(text to image, image to image, inpaint) and
+inputs (seed, reference images, negative prompt) it advertises, its
+request limits, and the size and image count the provider is asked for
+by default. The Generate panel and the Canvas only offer what the
+record advertises, so keep it in step with the provider's documentation.
+A model installed from the starter catalog keeps the catalog's
+capabilities and defaults, so those are shown read-only; its provider
+and provider model ID remain editable.
+
 :::note[Gated Downloads]
    The API Keys section is also used to provide keys for services that
    require authentication to download certain models, such as those

diff --git a/docs/src/content/docs/Users Guide/03.Models/External Models/gemini.mdx b/docs/src/content/docs/Users Guide/03.Models/External Models/gemini.mdx
@@ -39,6 +39,15 @@ Reference-image input is used to condition generation but counts as txt2img —
 
 All Gemini models are single-image-per-request — batch size is fixed at 1. To generate multiple variations, queue multiple invocations.
 
+## Provider-Specific Options
+
+Selecting a Gemini model adds a **Provider options** section to the Generate panel:
+
+- **Temperature** — Override the sampling temperature (0–2; the provider default is 1).
+- **Thinking level** — `minimal` or `high`, or the provider default.
+
+Gemini 3 models list their resolution presets under **Provider size** in the Size section. Picking one sets the width and height to the preset and sends its named size (`1K`, `2K`, `4K`) to the API. The named size and the sampling options are recorded in the image metadata and restored by **Recall All** and **Remix**.
+
 ## Tips
 
 <Steps>

diff --git a/docs/src/content/docs/Users Guide/03.Models/External Models/openai.mdx b/docs/src/content/docs/Users Guide/03.Models/External Models/openai.mdx
@@ -44,12 +44,19 @@ Inpainting (mask-based editing) is not currently supported for any OpenAI model
 
 ## Provider-Specific Options
 
-For **GPT Image** models, Invoke surfaces two provider-specific options in the parameters panel:
+Selecting an OpenAI model adds a **Provider options** section to the Generate panel:
 
 - **Quality** — `low`, `medium`, `high`, or `auto`. Higher quality costs more and takes longer.
 - **Background** — `auto`, `transparent`, or `opaque`. Use `transparent` for PNG output with an alpha channel.
+- **Input fidelity** — `low` or `high`, offered only for models that take image inputs (img2img or reference images). Leave it on the provider default to let OpenAI choose.
 
-DALL·E 2 and DALL·E 3 do not expose these options.
+These are recorded in the image metadata, so **Recall All** and **Remix** restore them when the image's model is selected.
+
+The **Size** section lists the provider's fixed sizes per aspect ratio (**Provider size**). Any other size is mapped to the closest one the provider accepts.
+
+## Canvas
+
+GPT Image models that advertise `img2img` can generate from the Canvas: an empty bounding box runs text-to-image, and a filled one sends the composited bounding box as the init image. The provider edits the composite whole at the bounding box's own size, so the render size, denoising strength, control layers and regional guidance do not apply and are not offered. Inpainting and outpainting are not available for OpenAI models.
 
 ## How Requests Are Routed
 

diff --git a/docs/src/content/docs/Users Guide/03.Models/External Models/seedream.mdx b/docs/src/content/docs/Users Guide/03.Models/External Models/seedream.mdx
@@ -48,11 +48,17 @@ All Seedream models share the same aspect ratio set: `1:1`, `2:3`, `3:2`, `3:4`,
 
 ## Provider-Specific Options
 
-Seedream exposes two provider-specific toggles in the parameters panel:
+Selecting a Seedream model adds a **Provider options** section to the Generate panel with two toggles:
 
 - **Watermark** — When enabled, BytePlus adds a small watermark to the output. Off by default.
 - **Optimize Prompt** — When enabled, BytePlus rewrites your prompt server-side for better generation quality. Useful for short prompts; disable if you want the exact wording preserved.
 
+Both are recorded in the image metadata and restored by **Recall All** and **Remix**. The Size section lists the provider's 2K sizes per aspect ratio under **Provider size**.
+
+## Canvas
+
+Seedream models advertise `img2img`, so they can generate from the Canvas: an empty bounding box runs text-to-image, and a filled one sends the composited bounding box as the init image. The provider edits the composite whole at the bounding box's own size, so the render size, denoising strength, control layers and regional guidance do not apply and are not offered. Inpainting and outpainting are not available.
+
 **Seed** and **guidance scale** are not accepted by the 4.x / 5.x family.
 
 ## Reference Images

diff --git a/invokeai/frontend/webv2/performance/architecture-baseline.json b/invokeai/frontend/webv2/performance/architecture-baseline.json
@@ -2,15 +2,15 @@
   "build": {
     "launchpad": {
       "baseline": {
-        "brotliBytes": 794977,
+        "brotliBytes": 795865,
         "cssRawBytes": 2159,
         "fontRawBytes": 219480,
-        "gzipBytes": 793689,
+        "gzipBytes": 794662,
         "imageRawBytes": 44398,
-        "initialRawBytes": 2042114,
-        "largestAssetRawBytes": 749029,
+        "initialRawBytes": 2046539,
+        "largestAssetRawBytes": 752421,
         "otherAssetRawBytes": 0,
-        "ownedRawBytes": 98621,
+        "ownedRawBytes": 98845,
         "requestCount": 35,
         "scriptRequestCount": 19,
         "sourceOwners": [
@@ -346,15 +346,15 @@
     },
     "editor": {
       "baseline": {
-        "brotliBytes": 1171873,
+        "brotliBytes": 1175651,
         "cssRawBytes": 2159,
         "fontRawBytes": 219480,
-        "gzipBytes": 1165072,
+        "gzipBytes": 1168861,
         "imageRawBytes": 44398,
-        "initialRawBytes": 3208701,
-        "largestAssetRawBytes": 749029,
+        "initialRawBytes": 3222002,
+        "largestAssetRawBytes": 752421,
         "otherAssetRawBytes": 0,
-        "ownedRawBytes": 150716,
+        "ownedRawBytes": 151795,
         "requestCount": 72,
         "scriptRequestCount": 56,
         "sourceOwners": [
@@ -550,6 +550,7 @@
           "source:src/features/generation/core/componentCompatibility.ts",
           "source:src/features/generation/core/conditioningRebalance.ts",
           "source:src/features/generation/core/dynamicPrompts.ts",
+          "source:src/features/generation/core/externalProvider.ts",
           "source:src/features/generation/core/graph.ts",
           "source:src/features/generation/core/graphBuilder.ts",
           "source:src/features/generation/core/krea2Conditioning.ts",
@@ -1092,7 +1093,7 @@
       "source": "src/app/WorkbenchApp.tsx"
     }
   },
-  "capturedAt": "2026-10-09",
+  "capturedAt": "2026-10-11",
   "developmentInvalidation": {
     "platformUiBarrel": {
       "specifier": "@platform/ui",