diff --git a/.changeset/gemini-image-ga-model-ids-and-per-model-sizes.md b/.changeset/gemini-image-ga-model-ids-and-per-model-sizes.md
new file mode 100644
index 0000000000..13da4ccc4f
--- /dev/null
+++ b/.changeset/gemini-image-ga-model-ids-and-per-model-sizes.md
@@ -0,0 +1,37 @@
+---
+'@tanstack/ai-gemini': minor
+---
+
+Add the GA Gemini native image model ids and give each native image model its own size type.
+
+`gemini-3.1-flash-image-preview` and `gemini-3-pro-image-preview` were shut down on 2026-06-25 and now 404. Their GA replacements — `gemini-3.1-flash-image` and `gemini-3-pro-image` — are now the primary ids. The `-preview` ids remain in the model union as aliases so existing code keeps compiling; `gemini-2.5-flash-image` stays fully supported ahead of its 2026-10-02 shutdown.
+
+Sizes were a single flat union (`{8 ratios}_{1K|2K|4K}`) applied to every native model. Google documents four different sets, so each model now maps to its own:
+
+| model                                   | aspect ratios | resolutions                      |
+| --------------------------------------- | ------------- | -------------------------------- |
+| `gemini-3.1-flash-image` (+ `-preview`) | 14            | `512` `1K` `2K` `4K`             |
+| `gemini-3.1-flash-lite-image`           | 14            | `1K`                             |
+| `gemini-3-pro-image` (+ `-preview`)     | 10            | `1K` `2K` `4K`                   |
+| `gemini-2.5-flash-image`                | 10            | none — bare ratio, e.g. `'16:9'` |
+
+`4:5` and `5:4` are now accepted on every native model (Google lists them for all four; the old union omitted them). `9:21` is deliberately still rejected — it exists on Vertex/Cloud only and the Gemini API rejects it.
+
+**Runtime behaviour changes in two places.** The rest of the change is types-only, but these two are real wire-format deltas:
+
+- `parseNativeImageSize()` now accepts a bare aspect ratio. Previously `'16:9'` failed to parse, so the adapter omitted `imageConfig` entirely and the model picked its own aspect ratio; it now parses to `{ aspectRatio: '16:9' }` and the adapter sends `imageConfig.aspectRatio = '16:9'`. A JavaScript caller — or a TypeScript caller whose `size` is computed at runtime and widened to `string` — that already passed a bare ratio will get a differently-framed image after upgrading, with no compile or runtime error.
+- Migrating a `gemini-2.5-flash-image` call from `'16:9_1K'` to the now-required bare `'16:9'` drops `imageSize` from the `generateContent` request. That is intended: Google publishes no `image_size` value or default for this model, so the adapter no longer guesses a tier the API never documented.
+
+**BREAKING (types only):** size combinations the selected model never supported no longer compile. No model id was removed.
+
+- `gemini-3.1-flash-lite-image`: `2K` and `4K` are rejected (the model only emits 1K). Use `'<ratio>_1K'`.
+- `gemini-3-pro-image` / `gemini-3-pro-image-preview`: the extreme banner ratios `1:4` `4:1` `1:8` `8:1` are rejected (Gemini 3.1 Flash Image only), as is the `512` tier.
+- `gemini-2.5-flash-image`: any `_1K` / `_2K` / `_4K` suffix is rejected — pass the bare ratio (`'16:9'`, not `'16:9_1K'`) — as are the four extreme banner ratios.
+- `GeminiNativeImageSize` is now the union of the per-model types rather than one flat template literal. It was not previously reachable from the package entry point, so this is a new export rather than a changed one.
+
+**New type exports**, so the per-model narrowing is nameable and not just inferred at the call site: `GeminiImageModelSizeByName`, `GeminiStandardImageAspectRatio`, `GeminiExtendedImageAspectRatio`, `Gemini31FlashImageSize`, `Gemini31FlashLiteImageSize`, `Gemini3ProImageSize`, `Gemini25FlashImageSize`, `GeminiNativeImageSize`.
+
+**Two caveats worth knowing before you rely on this.**
+
+- _No in-editor deprecation warning on the dead `-preview` ids._ The `@deprecated` tags live on module-private model-metadata consts, and `GeminiImageModels` is projected out of a const array (`(typeof GEMINI_IMAGE_MODELS)[number]`), which collapses to bare string literals — JSDoc does not survive that projection. So `geminiImage('gemini-3-pro-image-preview')` still compiles cleanly with no strikethrough and no hint, and fails only at request time. Grep your codebase for `-image-preview` rather than expecting the compiler to flag it.
+- _`gemini-3.1-flash-lite-image`'s four extreme ratios (`1:4` `4:1` `1:8` `8:1`) are partially inferred._ Unlike the other three native models, Flash Lite has no per-model ratio table on the Gemini API guide. The 14-value set rests on the Cloud model page's explicit enumeration plus `ai.google.dev`'s bare "a discrete set of 14 aspect ratios" assertion; the only Gemini-API enumeration for this model is a 10-item bullet prefixed "New aspect ratios", read here as a what's-new list rather than an exhaustive set. If the API rejects those four in practice, this type over-accepts and should narrow to the 10-ratio set.
diff --git a/docs/adapters/gemini.md b/docs/adapters/gemini.md
index 31da9cf6e2..54298b152d 100644
--- a/docs/adapters/gemini.md
+++ b/docs/adapters/gemini.md
@@ -411,7 +411,7 @@ See the [Embeddings guide](../embeddings.md) for the full API.
 
 The Gemini adapter supports two types of image generation:
 
-- **Gemini native image models** (NanoBanana) — Use the `generateContent` API with models like `gemini-3.1-flash-image-preview`. These support extended resolution tiers (1K, 2K, 4K) and aspect ratio control.
+- **Gemini native image models** (NanoBanana) — Use the `generateContent` API with models like `gemini-3.1-flash-image`. These support aspect ratio control plus resolution tiers (`512`, `1K`, `2K`, `4K`); which ratios and tiers are accepted varies per model and is enforced at compile time.
 - **Imagen models** — Use the `generateImages` API with models like `imagen-4.0-generate-001`. These are dedicated image generation models with WIDTHxHEIGHT sizing.
 
 The adapter automatically routes to the correct API based on the model name — models starting with `gemini-` use `generateContent`, while `imagen-` models use `generateImages`.
@@ -425,7 +425,7 @@ import { generateImage } from "@tanstack/ai";
 import { geminiImage } from "@tanstack/ai-gemini";
 
 const result = await generateImage({
-  adapter: geminiImage("gemini-3.1-flash-image-preview"),
+  adapter: geminiImage("gemini-3.1-flash-image"),
   prompt: "A futuristic cityscape at sunset",
   numberOfImages: 1,
   size: "16:9_4K",
@@ -462,10 +462,18 @@ size: "1:1_2K"
 size: "9:16_1K"
 ```
 
-| Component | Values |
-|-----------|--------|
-| Aspect Ratio | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `9:16`, `16:9`, `21:9` |
-| Resolution | `1K`, `2K`, `4K` |
+The accepted set differs per model, and each model's `size` is narrowed to its own set at compile time:
+
+| Model | Aspect ratios | Resolutions |
+|-------|---------------|-------------|
+| `gemini-3.1-flash-image` | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`, `1:4`, `4:1`, `1:8`, `8:1` | `512`, `1K`, `2K`, `4K` |
+| `gemini-3.1-flash-lite-image` | same 14 as above (see note) | `1K` only |
+| `gemini-3-pro-image` | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9` | `1K`, `2K`, `4K` |
+| `gemini-2.5-flash-image` | same 10 as above | none — pass the bare ratio, e.g. `size: "16:9"` |
+
+The `K` is case-sensitive (`1k` is rejected by the API), the smallest tier's token is `512` (not `512px` or `0.5K`), and `9:21` is Vertex/Cloud-only so it is not accepted here. Google documents no `image_size` for `gemini-2.5-flash-image`, so that model takes a bare aspect ratio and the adapter sends no `imageSize`.
+
+> **Note on `gemini-3.1-flash-lite-image`.** The four extreme banner ratios (`1:4`, `4:1`, `1:8`, `8:1`) are partially inferred for this model. Unlike the other three, Flash Lite has no per-model ratio table in Google's Gemini API guide; the 14-value set comes from the Cloud model page's explicit enumeration plus the guide's bare "a discrete set of 14 aspect ratios" assertion. The only Gemini-API enumeration for it is a 10-item bullet headed "New aspect ratios", which we read as a what's-new list rather than an exhaustive one. If the API rejects those four in practice, prefer the 10 standard ratios on this model.
 
 #### Imagen Models
 
@@ -547,14 +555,16 @@ GOOGLE_API_KEY=your-api-key-here
 
 ### Gemini Native Image Models (NanoBanana)
 
-These models use the `generateContent` API and support resolution tiers (1K, 2K, 4K).
+These models use the `generateContent` API and support per-model resolution tiers.
 
 | Model | Description |
 |-------|-------------|
-| `gemini-3.1-flash-image-preview` | Latest and fastest Gemini native image generation |
-| `gemini-3.1-flash-lite-image` | Nano Banana 2 Lite — ultra-low-latency, low-cost image generation |
-| `gemini-3-pro-image-preview` | Higher quality Gemini native image generation |
-| `gemini-2.5-flash-image` | Gemini 2.5 Flash with image generation |
+| `gemini-3.1-flash-image` | Nano Banana 2 — latest and fastest Gemini native image generation (512/1K/2K/4K) |
+| `gemini-3.1-flash-lite-image` | Nano Banana 2 Lite — ultra-low-latency, low-cost image generation (1K only) |
+| `gemini-3-pro-image` | Nano Banana Pro — higher quality Gemini native image generation (1K/2K/4K) |
+| `gemini-2.5-flash-image` | Nano Banana — legacy; shuts down 2026-10-02. Takes a bare aspect ratio |
+
+The `gemini-3.1-flash-image-preview` and `gemini-3-pro-image-preview` ids were shut down on 2026-06-25. They remain in the type union as deprecated aliases so existing code compiles, but calls to them fail — use the GA ids above.
 
 ### Imagen Models
 
diff --git a/docs/advanced/runtime-adapter-switching.md b/docs/advanced/runtime-adapter-switching.md
index e5c6189c98..8ac2cf25c1 100644
--- a/docs/advanced/runtime-adapter-switching.md
+++ b/docs/advanced/runtime-adapter-switching.md
@@ -116,7 +116,7 @@ export const Route = createFileRoute('/api/chat')({
 
 ## Using with Image Adapters
 
-The same pattern works for image generation:
+The same pattern works for image generation. Unlike the text and summarize adapters above, image adapters don't all accept the same shape of `size` — so it travels alongside its adapter in the provider map instead of being passed once for every branch:
 
 ```typescript
 import { generateImage } from '@tanstack/ai'
@@ -125,25 +125,28 @@ import { geminiImage } from '@tanstack/ai-gemini'
 
 type ImageProvider = 'openai' | 'gemini'
 
-const imageAdapters: Record<ImageProvider, () => ReturnType<typeof openaiImage | typeof geminiImage>> = {
-  openai: () => openaiImage('gpt-image-2'),
-  gemini: () => geminiImage('gemini-3.1-flash-image-preview'),
+const imageAdapters = {
+  openai: () => ({ adapter: openaiImage('gpt-image-2'), size: '1024x1024' as const }),
+  gemini: () => ({ adapter: geminiImage('gemini-3.1-flash-image'), size: '16:9_4K' as const }),
 }
 
 export async function POST(request: Request) {
   const body = await request.json()
   const provider: ImageProvider = body.provider ?? 'openai'
+  const { adapter, size } = imageAdapters[provider]()
 
   const result = await generateImage({
-    adapter: imageAdapters[provider](),
+    adapter,
     prompt: 'A beautiful sunset over mountains',
-    size: '1024x1024',
+    size,
   })
 
   return Response.json(result)
 }
 ```
 
+`size` is provider-specific, which is why it can't be a single literal shared across branches: Gemini native image models take a `'<aspectRatio>_<tier>'` string (e.g. `'16:9_4K'`), while OpenAI and Imagen models take pixel dimensions (e.g. `'1024x1024'`).
+
 ## Using with Summarize Adapters
 
 And for summarization:
diff --git a/docs/config.json b/docs/config.json
index e5c98bc877..3789654fb3 100644
--- a/docs/config.json
+++ b/docs/config.json
@@ -437,7 +437,7 @@
           "label": "Image Generation",
           "to": "media/image-generation",
           "addedAt": "2026-04-15",
-          "updatedAt": "2026-08-13"
+          "updatedAt": "2026-08-14"
         },
         {
           "label": "Video Generation",
@@ -671,7 +671,8 @@
         {
           "label": "Runtime Adapter Switching",
           "to": "advanced/runtime-adapter-switching",
-          "addedAt": "2026-04-15"
+          "addedAt": "2026-04-15",
+          "updatedAt": "2026-08-14"
         },
         {
           "label": "Tree-Shaking",
@@ -796,7 +797,7 @@
           "label": "Google Gemini",
           "to": "adapters/gemini",
           "addedAt": "2026-04-15",
-          "updatedAt": "2026-07-22"
+          "updatedAt": "2026-08-14"
         },
         {
           "label": "Ollama",
diff --git a/docs/media/image-generation.md b/docs/media/image-generation.md
index c1312e8122..6fed5f85b3 100644
--- a/docs/media/image-generation.md
+++ b/docs/media/image-generation.md
@@ -54,7 +54,7 @@ import { geminiImage } from "@tanstack/ai-gemini";
 
 // Gemini native model (NanoBanana) — uses generateContent API
 const result = await generateImage({
-  adapter: geminiImage("gemini-3.1-flash-image-preview"),
+  adapter: geminiImage("gemini-3.1-flash-image"),
   prompt: "A futuristic cityscape at night",
   size: "16:9_4K",
 });
@@ -121,19 +121,27 @@ All image adapters support these common options:
 
 #### Gemini Native Models (NanoBanana)
 
-Gemini native image models use a template literal size format: `"aspectRatio_resolution"`.
+Gemini native image models use a template literal size format: `"aspectRatio_resolution"`. Each model accepts its own set, narrowed at compile time:
 
-| Aspect Ratios                                             | Resolutions      |
-| --------------------------------------------------------- | ---------------- |
-| `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `9:16`, `16:9`, `21:9` | `1K`, `2K`, `4K` |
+| Model                         | Aspect Ratios                                                                                          | Resolutions             |
+| ----------------------------- | ------------------------------------------------------------------------------------------------------ | ----------------------- |
+| `gemini-3.1-flash-image`      | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`, `1:4`, `4:1`, `1:8`, `8:1` | `512`, `1K`, `2K`, `4K` |
+| `gemini-3.1-flash-lite-image` | same 14 as above (see note)                                                                             | `1K`                    |
+| `gemini-3-pro-image`          | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`                            | `1K`, `2K`, `4K`        |
+| `gemini-2.5-flash-image`      | same 10 as above                                                                                        | none — bare ratio       |
 
 ```typescript ignore
 // Examples
 size: "16:9_4K"; // Widescreen at 4K resolution
 size: "1:1_2K"; // Square at 2K resolution
-size: "9:16_1K"; // Portrait at 1K resolution
+size: "1:8_512"; // Tall banner at the 512 (0.5K) tier — Flash Image only
+size: "16:9"; // gemini-2.5-flash-image: bare ratio, no resolution suffix
 ```
 
+The `K` is case-sensitive (`1k` is rejected by the API), the smallest tier's token is `512` (not `512px` or `0.5K`), and `9:21` is Vertex/Cloud-only so it is not accepted. Google documents no `image_size` for `gemini-2.5-flash-image`, so that model takes a bare aspect ratio and the adapter sends no `imageSize`.
+
+> **Note on `gemini-3.1-flash-lite-image`.** The four extreme banner ratios (`1:4`, `4:1`, `1:8`, `8:1`) are partially inferred for this model. Unlike the other three, Flash Lite has no per-model ratio table in Google's Gemini API guide; the 14-value set comes from the Cloud model page's explicit enumeration plus the guide's bare "a discrete set of 14 aspect ratios" assertion. The only Gemini-API enumeration for it is a 10-item bullet headed "New aspect ratios", which we read as a what's-new list rather than an exhaustive one. If the API rejects those four in practice, prefer the 10 standard ratios on this model.
+
 #### Gemini Imagen Models
 
 Imagen models accept WIDTHxHEIGHT format, which maps to aspect ratios internally:
@@ -191,7 +199,7 @@ import { geminiImage } from "@tanstack/ai-gemini";
 import { badExampleUrl, goodExampleUrl } from "./urls";
 
 await generateImage({
-  adapter: geminiImage("gemini-3.1-flash-image-preview"),
+  adapter: geminiImage("gemini-3.1-flash-image"),
   prompt: [
     { type: "text", content: "Not like this" },
     { type: "image", source: { type: "url", value: badExampleUrl } },
@@ -334,7 +342,7 @@ import { generateImage } from "@tanstack/ai";
 import { geminiImage } from "@tanstack/ai-gemini";
 
 await generateImage({
-  adapter: geminiImage("gemini-3.1-flash-image-preview"),
+  adapter: geminiImage("gemini-3.1-flash-image"),
   prompt: [
     {
       type: "text",
@@ -641,7 +649,7 @@ import { generateImage } from "@tanstack/ai";
 import { geminiImage } from "@tanstack/ai-gemini";
 
 const result = await generateImage({
-  adapter: geminiImage("gemini-3.1-flash-image-preview"),
+  adapter: geminiImage("gemini-3.1-flash-image"),
   prompt: "A beautiful garden",
   size: "16:9_4K",
 });
@@ -709,12 +717,14 @@ if (result.usage?.unitsBilled != null) {
 
 #### Gemini Native Models (NanoBanana)
 
-| Model                            | Description                                                       |
-| -------------------------------- | ----------------------------------------------------------------- |
-| `gemini-3.1-flash-image-preview` | Latest and fastest Gemini native image generation                 |
-| `gemini-3.1-flash-lite-image`    | Nano Banana 2 Lite — ultra-low-latency, low-cost image generation |
-| `gemini-3-pro-image-preview`     | Higher quality Gemini native image generation                     |
-| `gemini-2.5-flash-image`         | Gemini 2.5 Flash with image generation                            |
+| Model                         | Description                                                                |
+| ----------------------------- | -------------------------------------------------------------------------- |
+| `gemini-3.1-flash-image`      | Nano Banana 2 — latest and fastest Gemini native image generation          |
+| `gemini-3.1-flash-lite-image` | Nano Banana 2 Lite — ultra-low-latency, low-cost image generation          |
+| `gemini-3-pro-image`          | Nano Banana Pro — higher quality Gemini native image generation            |
+| `gemini-2.5-flash-image`      | Nano Banana — legacy; shuts down 2026-10-02                                |
+
+The `gemini-3.1-flash-image-preview` and `gemini-3-pro-image-preview` ids were shut down on 2026-06-25. They remain in the type union as deprecated aliases so existing code compiles, but calls to them fail — use the GA ids above.
 
 #### Gemini Imagen Models
 
diff --git a/examples/ts-react-chat/src/routes/image-gen.tsx b/examples/ts-react-chat/src/routes/image-gen.tsx
index c891fd1464..d106f65f49 100644
--- a/examples/ts-react-chat/src/routes/image-gen.tsx
+++ b/examples/ts-react-chat/src/routes/image-gen.tsx
@@ -27,12 +27,12 @@ const IMAGE_MODELS = [
     label: 'Gemini 2.5 Flash Image',
   },
   {
-    value: 'google/gemini-2.5-flash-image-preview',
-    label: 'Gemini 2.5 Flash Image Preview',
+    value: 'google/gemini-3.1-flash-image',
+    label: 'Gemini 3.1 Flash Image (Nano Banana 2)',
   },
   {
-    value: 'google/gemini-3-pro-image-preview',
-    label: 'Gemini 3 Pro Image Preview',
+    value: 'google/gemini-3-pro-image',
+    label: 'Gemini 3 Pro Image (Nano Banana Pro)',
   },
 ] as const
 
diff --git a/examples/ts-react-media/src/lib/models.ts b/examples/ts-react-media/src/lib/models.ts
index 633f8da8bd..8ab505947c 100644
--- a/examples/ts-react-media/src/lib/models.ts
+++ b/examples/ts-react-media/src/lib/models.ts
@@ -48,7 +48,7 @@ export const IMAGE_MODELS = [
     provider: 'fal' as const,
   },
   {
-    id: 'gemini-3.1-flash-image-preview',
+    id: 'gemini-3.1-flash-image',
     name: 'NanoBanana 2 (Gemini 3.1 Flash)',
     description: 'Latest and fastest Gemini native image generation',
     defaultSize: '16:9_4K' as const,
@@ -56,7 +56,7 @@ export const IMAGE_MODELS = [
     provider: 'gemini' as const,
   },
   {
-    id: 'gemini-3-pro-image-preview',
+    id: 'gemini-3-pro-image',
     name: 'NanoBanana Pro (Gemini 3 Pro)',
     description: 'Higher quality Gemini native image generation',
     defaultSize: '16:9_4K' as const,
diff --git a/examples/ts-react-media/src/lib/server-functions.ts b/examples/ts-react-media/src/lib/server-functions.ts
index aa0f12c07a..717f3e1c47 100644
--- a/examples/ts-react-media/src/lib/server-functions.ts
+++ b/examples/ts-react-media/src/lib/server-functions.ts
@@ -196,17 +196,17 @@ export const generateImageFn = createServerFn({ method: 'POST' })
           },
         })
       }
-      case 'gemini-3.1-flash-image-preview': {
+      case 'gemini-3.1-flash-image': {
         return generateImage({
-          adapter: geminiImage('gemini-3.1-flash-image-preview'),
+          adapter: geminiImage('gemini-3.1-flash-image'),
           prompt: asImagePrompt(data.prompt),
           numberOfImages: 1,
           size: '16:9_4K',
         })
       }
-      case 'gemini-3-pro-image-preview': {
+      case 'gemini-3-pro-image': {
         return generateImage({
-          adapter: geminiImage('gemini-3-pro-image-preview'),
+          adapter: geminiImage('gemini-3-pro-image'),
           prompt: asImagePrompt(data.prompt),
           numberOfImages: 1,
           size: '16:9_4K',
diff --git a/packages/ai-gemini/src/image/image-provider-options.ts b/packages/ai-gemini/src/image/image-provider-options.ts
index 84fd66da76..951a60993e 100644
--- a/packages/ai-gemini/src/image/image-provider-options.ts
+++ b/packages/ai-gemini/src/image/image-provider-options.ts
@@ -145,47 +145,119 @@ export type GeminiImageSize =
   | '1080x1920'
 
 /**
- * Aspect ratios supported by Gemini native image models (via generateContent API).
- * Matches the SDK's ImageConfig.aspectRatio values.
+ * The ten aspect ratios every Gemini native image model accepts.
+ *
+ * Note `9:21` is deliberately absent: it exists only on Vertex / Cloud and is
+ * rejected by the Gemini API (`generateContent`), which is the surface this
+ * adapter targets.
+ *
+ * @see https://ai.google.dev/gemini-api/docs/image-generation
  */
-export type GeminiNativeImageAspectRatio =
+export type GeminiStandardImageAspectRatio =
   | '1:1'
   | '2:3'
   | '3:2'
   | '3:4'
   | '4:3'
+  | '4:5'
+  | '5:4'
   | '9:16'
   | '16:9'
   | '21:9'
 
 /**
- * Resolution tiers for Gemini native image models.
- * Matches the SDK's ImageConfig.imageSize values.
+ * The ten standard ratios plus the four extreme banner/strip ratios that only
+ * the Gemini 3.1 Flash Image models accept — 14 values, matching the
+ * `generateContent` `ImageConfig.aspectRatio` field union.
+ *
+ * @see https://ai.google.dev/api/generate-content
+ */
+export type GeminiExtendedImageAspectRatio =
+  | GeminiStandardImageAspectRatio
+  | '1:4'
+  | '4:1'
+  | '1:8'
+  | '8:1'
+
+/**
+ * Sizes for `gemini-3.1-flash-image` (and its shut-down `-preview` alias):
+ * all 14 aspect ratios at 512 / 1K / 2K / 4K. `512` is the wire token for the
+ * 0.5K tier — not `512px`, and the `K` is case-sensitive (`1k` is rejected).
+ */
+export type Gemini31FlashImageSize =
+  `${GeminiExtendedImageAspectRatio}_${'512' | '1K' | '2K' | '4K'}`
+
+/**
+ * Sizes for `gemini-3.1-flash-lite-image`: all 14 aspect ratios, 1K only —
+ * 2K and 4K are explicitly unsupported on this model.
+ *
+ * Caveat on the ratio set: the four extreme banner ratios (`1:4`, `4:1`,
+ * `1:8`, `8:1`) are the one partially-inferred cell in this file. Unlike the
+ * other three native models, Flash Lite has no per-model ratio table on the
+ * Gemini API guide. The 14 rest on the Cloud model page's explicit 14-value
+ * enumeration plus `ai.google.dev`'s bare assertion that the model "supports a
+ * discrete set of 14 aspect ratios"; the only Gemini-API enumeration for it is
+ * a 10-item bullet prefixed "New aspect ratios", read here as a what's-new
+ * list rather than an exhaustive set. If a live probe shows the extremes are
+ * rejected, narrow this to {@link GeminiStandardImageAspectRatio}.
+ *
+ * @see https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-1-flash-lite-image
+ * @see https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-lite-image
+ */
+export type Gemini31FlashLiteImageSize = `${GeminiExtendedImageAspectRatio}_1K`
+
+/**
+ * Sizes for `gemini-3-pro-image` (and its shut-down `-preview` alias): the ten
+ * standard aspect ratios at 1K / 2K / 4K. Pro has no 512 tier and none of the
+ * extreme banner ratios on the Gemini API.
+ */
+export type Gemini3ProImageSize =
+  `${GeminiStandardImageAspectRatio}_${'1K' | '2K' | '4K'}`
+
+/**
+ * Sizes for `gemini-2.5-flash-image`: a bare aspect ratio with no resolution
+ * suffix, e.g. `'16:9'`. Google documents no `image_size` value or default for
+ * this model — it emits a single fixed 1024px-class output — so the adapter
+ * sends `imageConfig.aspectRatio` and omits `imageSize` entirely rather than
+ * guessing a tier the API never documented.
  */
-export type GeminiNativeImageResolution = '1K' | '2K' | '4K'
+export type Gemini25FlashImageSize = GeminiStandardImageAspectRatio
 
 /**
- * Template literal size type for Gemini native image models: "16:9_4K", "1:1_2K", etc.
+ * Any size accepted by any Gemini native image model. Prefer the per-model
+ * narrowing in {@link GeminiImageModelSizeByName} — this union is the widest
+ * possible set and accepts combinations no single model supports.
  */
 export type GeminiNativeImageSize =
-  `${GeminiNativeImageAspectRatio}_${GeminiNativeImageResolution}`
+  | Gemini31FlashImageSize
+  | Gemini31FlashLiteImageSize
+  | Gemini3ProImageSize
+  | Gemini25FlashImageSize
 
 /**
  * Gemini native image models that use the generateContent API path.
- * These models support template literal sizes (aspectRatio_resolution).
+ * These models take an aspect-ratio-based size rather than Imagen's
+ * WIDTHxHEIGHT pixel strings.
  */
 export type GeminiNativeImageModels =
+  | 'gemini-3.1-flash-image'
   | 'gemini-3.1-flash-image-preview'
   | 'gemini-3.1-flash-lite-image'
+  | 'gemini-3-pro-image'
   | 'gemini-3-pro-image-preview'
   | 'gemini-2.5-flash-image'
 
 /**
- * Model-specific size options mapping.
- * Gemini native image models use template literal sizes, Imagen models use pixel sizes.
+ * Model-specific size options mapping. Each native model gets its own ratio ×
+ * resolution set (they genuinely differ); Imagen models use pixel sizes.
  */
 export type GeminiImageModelSizeByName = {
-  [K in GeminiNativeImageModels]: GeminiNativeImageSize
+  'gemini-3.1-flash-image': Gemini31FlashImageSize
+  'gemini-3.1-flash-image-preview': Gemini31FlashImageSize
+  'gemini-3.1-flash-lite-image': Gemini31FlashLiteImageSize
+  'gemini-3-pro-image': Gemini3ProImageSize
+  'gemini-3-pro-image-preview': Gemini3ProImageSize
+  'gemini-2.5-flash-image': Gemini25FlashImageSize
 } & {
   [K in Exclude<GeminiImageModels, GeminiNativeImageModels>]: GeminiImageSize
 }
@@ -309,13 +381,23 @@ export function validatePrompt(options: {
 
 /**
  * Parses a Gemini native image size string into its components.
- * Format: "aspectRatio_resolution" e.g. "16:9_4K" → { aspectRatio: "16:9", resolution: "4K" }
+ *
+ * Format: `"aspectRatio_resolution"`, e.g. `"16:9_4K"` →
+ * `{ aspectRatio: "16:9", resolution: "4K" }`.
+ *
+ * The resolution suffix is optional: `gemini-2.5-flash-image` takes a bare
+ * aspect ratio (`"16:9"` → `{ aspectRatio: "16:9" }`) because Google documents
+ * no `image_size` for it, and the caller must then omit `imageSize` from the
+ * request rather than substituting a default.
  */
 export function parseNativeImageSize(
   size: string,
-): { aspectRatio: string; resolution: string } | undefined {
-  const match = size.match(/^(\d+:\d+)_(.+)$/)
+): { aspectRatio: string; resolution?: string } | undefined {
+  const match = size.match(/^(\d+:\d+)(?:_(.+))?$/)
   const [, aspectRatio, resolution] = match ?? []
-  if (aspectRatio === undefined || resolution === undefined) return undefined
-  return { aspectRatio, resolution }
+  if (aspectRatio === undefined) return undefined
+  return {
+    aspectRatio,
+    ...(resolution !== undefined && { resolution }),
+  }
 }
diff --git a/packages/ai-gemini/src/index.ts b/packages/ai-gemini/src/index.ts
index e0c58922c0..f7ed26ceab 100644
--- a/packages/ai-gemini/src/index.ts
+++ b/packages/ai-gemini/src/index.ts
@@ -30,6 +30,18 @@ export type {
   GeminiImageProviderOptions,
   GeminiImageModelProviderOptionsByName,
   GeminiAspectRatio,
+  // Per-model size narrowing. `GeminiImageModelSizeByName` is the map
+  // `generateImage()` applies at the call site; the per-model aliases let you
+  // name a single model's set directly. `GeminiNativeImageSize` is the widest
+  // union across all native models — prefer the narrower types above it.
+  GeminiImageModelSizeByName,
+  GeminiStandardImageAspectRatio,
+  GeminiExtendedImageAspectRatio,
+  Gemini31FlashImageSize,
+  Gemini31FlashLiteImageSize,
+  Gemini3ProImageSize,
+  Gemini25FlashImageSize,
+  GeminiNativeImageSize,
   // Re-export SDK types for convenience
   PersonGeneration,
   SafetyFilterLevel,
diff --git a/packages/ai-gemini/src/model-meta.ts b/packages/ai-gemini/src/model-meta.ts
index 1ec09c5a86..aae4957c5f 100644
--- a/packages/ai-gemini/src/model-meta.ts
+++ b/packages/ai-gemini/src/model-meta.ts
@@ -118,7 +118,46 @@ const GEMINI_3_FLASH = {
     GeminiThinkingOptions
 >
 
+/**
+ * Gemini 3 Pro Image ("Nano Banana Pro") — GA. Accepts the ten standard
+ * aspect ratios at 1K / 2K / 4K.
+ * @see https://ai.google.dev/gemini-api/docs/models/gemini-3-pro-image
+ */
 const GEMINI_3_PRO_IMAGE = {
+  name: 'gemini-3-pro-image',
+  max_input_tokens: 65_536,
+  max_output_tokens: 32_768,
+  knowledge_cutoff: '2025-01-01',
+  supports: {
+    input: ['text', 'image'],
+    output: ['text', 'image'],
+    capabilities: ['batch_api', 'structured_output', 'thinking'],
+    tools: ['google_search'],
+  },
+  pricing: {
+    input: {
+      normal: 2,
+    },
+    output: {
+      normal: 0.134,
+    },
+  },
+} as const satisfies ModelMeta<
+  GeminiToolConfigOptions &
+    GeminiSafetyOptions &
+    GeminiCommonConfigOptions &
+    GeminiCachedContentOptions &
+    GeminiStructuredOutputOptions &
+    GeminiThinkingOptions
+>
+
+/**
+ * @deprecated `gemini-3-pro-image-preview` was shut down on 2026-06-25. Use
+ * the GA id `gemini-3-pro-image` instead — the preview id now 404s.
+ * Kept in the model union so existing code still compiles.
+ * @see https://ai.google.dev/gemini-api/docs/deprecations
+ */
+const GEMINI_3_PRO_IMAGE_PREVIEW = {
   name: 'gemini-3-pro-image-preview',
   max_input_tokens: 65_536,
   max_output_tokens: 32_768,
@@ -146,7 +185,47 @@ const GEMINI_3_PRO_IMAGE = {
     GeminiThinkingOptions
 >
 
+/**
+ * Gemini 3.1 Flash Image ("Nano Banana 2") — GA. The only native image model
+ * that accepts the four extreme banner ratios (1:4, 4:1, 1:8, 8:1) and the
+ * 512 (0.5K) resolution tier.
+ * @see https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-image
+ */
 const GEMINI_3_1_FLASH_IMAGE = {
+  name: 'gemini-3.1-flash-image',
+  max_input_tokens: 65_536,
+  max_output_tokens: 65_536,
+  knowledge_cutoff: '2025-01-01',
+  supports: {
+    input: ['text', 'image'],
+    output: ['text', 'image'],
+    capabilities: ['batch_api', 'structured_output', 'thinking'],
+    tools: ['google_search'],
+  },
+  pricing: {
+    input: {
+      normal: 0.25,
+    },
+    output: {
+      normal: 1.5,
+    },
+  },
+} as const satisfies ModelMeta<
+  GeminiToolConfigOptions &
+    GeminiSafetyOptions &
+    GeminiCommonConfigOptions &
+    GeminiCachedContentOptions &
+    GeminiStructuredOutputOptions &
+    GeminiThinkingOptions
+>
+
+/**
+ * @deprecated `gemini-3.1-flash-image-preview` was shut down on 2026-06-25.
+ * Use the GA id `gemini-3.1-flash-image` instead — the preview id now 404s.
+ * Kept in the model union so existing code still compiles.
+ * @see https://ai.google.dev/gemini-api/docs/deprecations
+ */
+const GEMINI_3_1_FLASH_IMAGE_PREVIEW = {
   name: 'gemini-3.1-flash-image-preview',
   max_input_tokens: 65_536,
   max_output_tokens: 65_536,
@@ -174,6 +253,10 @@ const GEMINI_3_1_FLASH_IMAGE = {
     GeminiThinkingOptions
 >
 
+/**
+ * Gemini 3.1 Flash Lite Image ("Nano Banana 2 Lite") — GA. 1K output only.
+ * @see https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-lite-image
+ */
 const GEMINI_3_1_FLASH_LITE_IMAGE = {
   name: 'gemini-3.1-flash-lite-image',
   max_input_tokens: 65_536,
@@ -375,6 +458,17 @@ const GEMINI_2_5_FLASH = {
     GeminiThinkingOptions
 >
 
+/**
+ * Gemini 2.5 Flash Image ("Nano Banana") — still GA, but documented as the
+ * legacy member of the family. Google publishes no `image_size` value for it,
+ * so its size type is a bare aspect ratio and the adapter sends no
+ * `imageConfig.imageSize`.
+ * @deprecated `gemini-2.5-flash-image` shuts down on 2026-10-02. Migrate to
+ * `gemini-3.1-flash-lite-image` (cheapest successor) or
+ * `gemini-3.1-flash-image`. Google's deprecations table still names the
+ * already-dead `gemini-3.1-flash-image-preview` as the replacement.
+ * @see https://ai.google.dev/gemini-api/docs/deprecations
+ */
 const GEMINI_2_5_FLASH_IMAGE = {
   name: 'gemini-2.5-flash-image',
   max_input_tokens: 1_048_576,
@@ -948,6 +1042,11 @@ export type GeminiModels = (typeof GEMINI_MODELS)[number]
 
 export type GeminiImageModels = (typeof GEMINI_IMAGE_MODELS)[number]
 
+/**
+ * Image generation models. GA ids come first; the trailing `-preview` ids are
+ * shut-down aliases kept only so existing code keeps compiling — new code
+ * should use the GA id above its alias.
+ */
 export const GEMINI_IMAGE_MODELS = [
   GEMINI_3_1_FLASH_IMAGE.name,
   GEMINI_3_1_FLASH_LITE_IMAGE.name,
@@ -956,6 +1055,9 @@ export const GEMINI_IMAGE_MODELS = [
   IMAGEN_4_GENERATE.name,
   IMAGEN_4_GENERATE_FAST.name,
   IMAGEN_4_GENERATE_ULTRA.name,
+  // Deprecated aliases — shut down 2026-06-25.
+  GEMINI_3_1_FLASH_IMAGE_PREVIEW.name,
+  GEMINI_3_PRO_IMAGE_PREVIEW.name,
 ] as const
 
 /**
diff --git a/packages/ai-gemini/tests/image-adapter.test.ts b/packages/ai-gemini/tests/image-adapter.test.ts
index 7f07bb04f0..d503c61d7e 100644
--- a/packages/ai-gemini/tests/image-adapter.test.ts
+++ b/packages/ai-gemini/tests/image-adapter.test.ts
@@ -152,11 +152,34 @@ describe('Gemini Image Adapter', () => {
       })
     })
 
+    it('parses the per-model sizes, including a bare ratio with no resolution', () => {
+      // 4:5 / 5:4 are documented for all four native models but were missing
+      // from the old shared 8-ratio union; "512" is the wire token for the
+      // 0.5K tier — not "512px", not "0.5K".
+      expect(parseNativeImageSize('4:5_2K')).toEqual({
+        aspectRatio: '4:5',
+        resolution: '2K',
+      })
+      expect(parseNativeImageSize('5:4_1K')).toEqual({
+        aspectRatio: '5:4',
+        resolution: '1K',
+      })
+      expect(parseNativeImageSize('1:8_512')).toEqual({
+        aspectRatio: '1:8',
+        resolution: '512',
+      })
+      // gemini-2.5-flash-image has no documented image_size, so its sizes are
+      // bare ratios: the resolution is absent, not defaulted.
+      expect(parseNativeImageSize('16:9')).toEqual({ aspectRatio: '16:9' })
+      expect(parseNativeImageSize('16:9')?.resolution).toBeUndefined()
+    })
+
     it('returns undefined for invalid formats', () => {
       expect(parseNativeImageSize('1024x1024')).toBeUndefined()
       expect(parseNativeImageSize('invalid')).toBeUndefined()
-      expect(parseNativeImageSize('16:9')).toBeUndefined()
       expect(parseNativeImageSize('4K')).toBeUndefined()
+      // A trailing separator with no resolution is not a bare ratio.
+      expect(parseNativeImageSize('16:9_')).toBeUndefined()
     })
   })
 
@@ -340,7 +363,8 @@ describe('Gemini Image Adapter', () => {
       const result = await generateImage({
         adapter,
         prompt: 'A red circle',
-        size: '1:1_2K',
+        // Flash Lite is 1K-only; 2K/4K are rejected at compile time.
+        size: '1:1_1K',
       })
 
       expect(mockGenerateContent).toHaveBeenCalledWith({
@@ -350,7 +374,7 @@ describe('Gemini Image Adapter', () => {
           responseModalities: ['TEXT', 'IMAGE'],
           imageConfig: {
             aspectRatio: '1:1',
-            imageSize: '2K',
+            imageSize: '1K',
           },
         },
       })
@@ -360,6 +384,173 @@ describe('Gemini Image Adapter', () => {
       expect(result.images[0]!.b64Json).toBe('lite-base64-image')
     })
 
+    it('routes the GA id gemini-3.1-flash-image through generateContent and sends the 512 tier', async () => {
+      const mockResponse = {
+        candidates: [
+          {
+            content: {
+              parts: [
+                {
+                  inlineData: {
+                    mimeType: 'image/png',
+                    data: 'ga-flash-image',
+                  },
+                },
+              ],
+            },
+          },
+        ],
+      }
+
+      const mockGenerateContent = vi.fn().mockResolvedValueOnce(mockResponse)
+      const mockGenerateImages = vi.fn()
+
+      const adapter = createGeminiImage(
+        'gemini-3.1-flash-image',
+        'test-api-key',
+      )
+      ;(
+        adapter as unknown as {
+          client: {
+            models: { generateContent: unknown; generateImages: unknown }
+          }
+        }
+      ).client = {
+        models: {
+          generateContent: mockGenerateContent,
+          generateImages: mockGenerateImages,
+        },
+      }
+
+      const result = await generateImage({
+        adapter,
+        prompt: 'A tall banner',
+        size: '1:8_512',
+      })
+
+      expect(mockGenerateContent).toHaveBeenCalledWith({
+        model: 'gemini-3.1-flash-image',
+        contents: 'A tall banner',
+        config: {
+          responseModalities: ['TEXT', 'IMAGE'],
+          imageConfig: {
+            aspectRatio: '1:8',
+            imageSize: '512',
+          },
+        },
+      })
+      // Native models must never take the Imagen generateImages path.
+      expect(mockGenerateImages).not.toHaveBeenCalled()
+      expect(result.images[0]!.b64Json).toBe('ga-flash-image')
+    })
+
+    it('routes the GA id gemini-3-pro-image through generateContent', async () => {
+      const mockResponse = {
+        candidates: [
+          {
+            content: {
+              parts: [
+                {
+                  inlineData: { mimeType: 'image/png', data: 'ga-pro-image' },
+                },
+              ],
+            },
+          },
+        ],
+      }
+
+      const mockGenerateContent = vi.fn().mockResolvedValueOnce(mockResponse)
+      const mockGenerateImages = vi.fn()
+
+      const adapter = createGeminiImage('gemini-3-pro-image', 'test-api-key')
+      ;(
+        adapter as unknown as {
+          client: {
+            models: { generateContent: unknown; generateImages: unknown }
+          }
+        }
+      ).client = {
+        models: {
+          generateContent: mockGenerateContent,
+          generateImages: mockGenerateImages,
+        },
+      }
+
+      const result = await generateImage({
+        adapter,
+        prompt: 'A portrait',
+        size: '4:5_2K',
+      })
+
+      expect(mockGenerateContent).toHaveBeenCalledWith({
+        model: 'gemini-3-pro-image',
+        contents: 'A portrait',
+        config: {
+          responseModalities: ['TEXT', 'IMAGE'],
+          imageConfig: {
+            aspectRatio: '4:5',
+            imageSize: '2K',
+          },
+        },
+      })
+      expect(mockGenerateImages).not.toHaveBeenCalled()
+      expect(result.images[0]!.b64Json).toBe('ga-pro-image')
+    })
+
+    it('sends aspectRatio but no imageSize for gemini-2.5-flash-image', async () => {
+      // Google documents no image_size for this model, so the adapter must
+      // send the ratio alone rather than inventing a resolution tier.
+      const mockResponse = {
+        candidates: [
+          {
+            content: {
+              parts: [
+                {
+                  inlineData: { mimeType: 'image/png', data: 'legacy-image' },
+                },
+              ],
+            },
+          },
+        ],
+      }
+
+      const mockGenerateContent = vi.fn().mockResolvedValueOnce(mockResponse)
+
+      const adapter = createGeminiImage(
+        'gemini-2.5-flash-image',
+        'test-api-key',
+      )
+      ;(
+        adapter as unknown as {
+          client: { models: { generateContent: unknown } }
+        }
+      ).client = {
+        models: {
+          generateContent: mockGenerateContent,
+        },
+      }
+
+      await generateImage({
+        adapter,
+        prompt: 'A wide landscape',
+        size: '16:9',
+      })
+
+      expect(mockGenerateContent).toHaveBeenCalledWith({
+        model: 'gemini-2.5-flash-image',
+        contents: 'A wide landscape',
+        config: {
+          responseModalities: ['TEXT', 'IMAGE'],
+          imageConfig: {
+            aspectRatio: '16:9',
+          },
+        },
+      })
+
+      const config = mockGenerateContent.mock.calls[0]![0].config
+      expect('imageSize' in config.imageConfig).toBe(false)
+    })
+
     it('surfaces token usage from usageMetadata (#330)', async () => {
       const mockResponse = {
         candidates: [
diff --git a/packages/ai-gemini/tests/image-per-model-type-safety.test.ts b/packages/ai-gemini/tests/image-per-model-type-safety.test.ts
new file mode 100644
index 0000000000..96d47a9e8e
--- /dev/null
+++ b/packages/ai-gemini/tests/image-per-model-type-safety.test.ts
@@ -0,0 +1,240 @@
+/**
+ * Per-model type-safety tests for Gemini `generateImage()` sizes.
+ *
+ * Positive cases: each documented (model, size) pair compiles cleanly.
+ * Negative cases: each undocumented pair produces a `@ts-expect-error`. `tsc`
+ * errors on an *unused* `@ts-expect-error`, so a green typecheck is itself the
+ * proof that every negative below is genuinely rejected.
+ *
+ * Companion to `chat-per-model-type-safety.test.ts` (modelOptions) and
+ * `tools-per-model-type-safety.test.ts` (tools). Compile-time only —
+ * `createImageOptions` is the identity helper for `generateImage()` options,
+ * so nothing here touches the network.
+ */
+import { describe, expectTypeOf, it } from 'vitest'
+import { createImageOptions } from '@tanstack/ai'
+import { createGeminiImage } from '../src/adapters/image'
+import type { GeminiImageModelSizeByName } from '../src/image/image-provider-options'
+
+const apiKey = 'test-api-key'
+
+describe('Gemini per-model image size gating', () => {
+  describe('gemini-3.1-flash-image — 14 ratios × 512/1K/2K/4K', () => {
+    it('accepts the extreme banner ratios and the 512 tier', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image', apiKey),
+        prompt: 'a wide banner',
+        size: '1:8_512',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image', apiKey),
+        prompt: 'a wide banner',
+        size: '8:1_4K',
+      })
+    })
+
+    it('accepts the flexible 4:5 / 5:4 ratios', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image', apiKey),
+        prompt: 'a portrait',
+        size: '4:5_1K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image', apiKey),
+        prompt: 'a landscape',
+        size: '5:4_2K',
+      })
+    })
+
+    it('rejects the Cloud-only 9:21 ratio and lowercase resolutions', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image', apiKey),
+        prompt: 'a tall strip',
+        // @ts-expect-error - 9:21 exists on Vertex/Cloud only; the Gemini API rejects it
+        size: '9:21_1K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image', apiKey),
+        prompt: 'a square',
+        // @ts-expect-error - the K is case-sensitive; "1k" is rejected by the API
+        size: '1:1_1k',
+      })
+    })
+
+    it('applies the same set to the shut-down -preview alias', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-image-preview', apiKey),
+        prompt: 'a wide banner',
+        size: '1:8_512',
+      })
+    })
+  })
+
+  describe('gemini-3.1-flash-lite-image — 14 ratios, 1K only', () => {
+    it('accepts 1K at any of the 14 ratios', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-lite-image', apiKey),
+        prompt: 'a square',
+        size: '1:1_1K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-lite-image', apiKey),
+        prompt: 'a banner',
+        size: '4:1_1K',
+      })
+    })
+
+    it('rejects every resolution above 1K', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-lite-image', apiKey),
+        prompt: 'a landscape',
+        // @ts-expect-error - Flash Lite Image only supports 1K; 2K/4K are unsupported
+        size: '16:9_4K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-lite-image', apiKey),
+        prompt: 'a landscape',
+        // @ts-expect-error - Flash Lite Image only supports 1K
+        size: '16:9_2K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3.1-flash-lite-image', apiKey),
+        prompt: 'a landscape',
+        // @ts-expect-error - the 512 tier is Gemini 3.1 Flash Image only
+        size: '16:9_512',
+      })
+    })
+  })
+
+  describe('gemini-3-pro-image — 10 ratios × 1K/2K/4K', () => {
+    it('accepts the flexible 4:5 ratio at 2K', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3-pro-image', apiKey),
+        prompt: 'a portrait',
+        size: '4:5_2K',
+      })
+    })
+
+    it('rejects the extreme banner ratios', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3-pro-image', apiKey),
+        prompt: 'a tall strip',
+        // @ts-expect-error - 1:8 is Gemini 3.1 Flash Image only; Pro takes the ten standard ratios
+        size: '1:8_1K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3-pro-image', apiKey),
+        prompt: 'a wide strip',
+        // @ts-expect-error - 4:1 is Gemini 3.1 Flash Image only
+        size: '4:1_2K',
+      })
+    })
+
+    it('rejects the 512 tier', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3-pro-image', apiKey),
+        prompt: 'a square',
+        // @ts-expect-error - Pro has no 512 (0.5K) tier
+        size: '1:1_512',
+      })
+    })
+
+    it('applies the same set to the shut-down -preview alias', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-3-pro-image-preview', apiKey),
+        prompt: 'a portrait',
+        size: '4:5_2K',
+      })
+    })
+  })
+
+  describe('gemini-2.5-flash-image — bare aspect ratio, no resolution', () => {
+    it('accepts a bare ratio', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-2.5-flash-image', apiKey),
+        prompt: 'a landscape',
+        size: '16:9',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-2.5-flash-image', apiKey),
+        prompt: 'a portrait',
+        size: '4:5',
+      })
+    })
+
+    it('rejects any resolution suffix', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-2.5-flash-image', apiKey),
+        prompt: 'a landscape',
+        // @ts-expect-error - Google documents no image_size for this model
+        size: '16:9_2K',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('gemini-2.5-flash-image', apiKey),
+        prompt: 'a landscape',
+        // @ts-expect-error - Google documents no image_size for this model
+        size: '16:9_1K',
+      })
+    })
+
+    it('rejects the extreme banner ratios', () => {
+      createImageOptions({
+        adapter: createGeminiImage('gemini-2.5-flash-image', apiKey),
+        prompt: 'a wide strip',
+        // @ts-expect-error - 8:1 is Gemini 3.1 Flash Image only
+        size: '8:1',
+      })
+    })
+  })
+
+  describe('Imagen models keep pixel sizes', () => {
+    it('accepts WIDTHxHEIGHT and rejects aspect-ratio sizes', () => {
+      createImageOptions({
+        adapter: createGeminiImage('imagen-4.0-generate-001', apiKey),
+        prompt: 'a cat',
+        size: '1024x1024',
+      })
+      createImageOptions({
+        adapter: createGeminiImage('imagen-4.0-generate-001', apiKey),
+        prompt: 'a cat',
+        // @ts-expect-error - Imagen takes pixel sizes, not aspect-ratio sizes
+        size: '16:9_2K',
+      })
+    })
+  })
+
+  describe('Model name type safety', () => {
+    it('rejects unknown image model names at the factory', () => {
+      // @ts-expect-error - 'gemini-9-pro-image' is not a Gemini image model
+      createGeminiImage('gemini-9-pro-image', apiKey)
+    })
+  })
+})
+
+describe('Gemini image size map shape assertions', () => {
+  it('maps each GA id to its own size set', () => {
+    expectTypeOf<
+      GeminiImageModelSizeByName['gemini-3.1-flash-image']
+    >().toEqualTypeOf<
+      GeminiImageModelSizeByName['gemini-3.1-flash-image-preview']
+    >()
+    expectTypeOf<
+      GeminiImageModelSizeByName['gemini-3-pro-image']
+    >().toEqualTypeOf<
+      GeminiImageModelSizeByName['gemini-3-pro-image-preview']
+    >()
+  })
+
+  it('gives the four native models four different size sets', () => {
+    expectTypeOf<
+      GeminiImageModelSizeByName['gemini-3.1-flash-image']
+    >().not.toEqualTypeOf<
+      GeminiImageModelSizeByName['gemini-3.1-flash-lite-image']
+    >()
+    expectTypeOf<
+      GeminiImageModelSizeByName['gemini-3-pro-image']
+    >().not.toEqualTypeOf<
+      GeminiImageModelSizeByName['gemini-2.5-flash-image']
+    >()
+  })
+})
diff --git a/packages/ai/skills/ai-core/adapter-configuration/references/gemini-adapter.md b/packages/ai/skills/ai-core/adapter-configuration/references/gemini-adapter.md
index f21c0b1ab5..fb27b468b1 100644
--- a/packages/ai/skills/ai-core/adapter-configuration/references/gemini-adapter.md
+++ b/packages/ai/skills/ai-core/adapter-configuration/references/gemini-adapter.md
@@ -94,8 +94,12 @@ Note: `GOOGLE_GENAI_API_KEY` does NOT work.
 ## Gotchas
 
 - All Gemini models are multimodal (text, image, audio, video, document input).
-- Image generation models (`gemini-3-pro-image-preview`, etc.) have smaller
-  input limits (65K tokens) compared to text models (1M tokens).
+- Image generation models (`gemini-3-pro-image`, `gemini-3.1-flash-image`, etc.)
+  have smaller input limits (65K tokens) compared to text models (1M tokens).
+- Use the GA image ids. `gemini-3-pro-image-preview` and
+  `gemini-3.1-flash-image-preview` were shut down on 2026-06-25 and now 404;
+  they remain in the type union only as deprecated aliases, so a call to them
+  compiles and then fails at runtime.
 - `thinkingConfig.thinkingLevel` (level-based) and `thinkingConfig.thinkingBudget`
   (budget-based) serve different models. Check which your model supports.
 - `cachedContent` must follow the format `cachedContents/{id}`.
diff --git a/packages/ai/skills/ai-core/media-generation/SKILL.md b/packages/ai/skills/ai-core/media-generation/SKILL.md
index fdf185a615..c7ca7e055d 100644
--- a/packages/ai/skills/ai-core/media-generation/SKILL.md
+++ b/packages/ai/skills/ai-core/media-generation/SKILL.md
@@ -151,9 +151,16 @@ function ImageGenerator() {
 ### 1. Image Generation
 
 Supported adapters: `openaiImage` (dall-e-2, dall-e-3, gpt-image-1,
-gpt-image-1-mini, gpt-image-2), `geminiImage` (gemini-3.1-flash-image-preview,
-gemini-3.1-flash-lite-image, imagen-4.0-generate-001, etc.) and `byteplusImage`
-(Seedream — `seedream-4-0-250828`, `seedream-4-5-251128`, the 5.0 family).
+gpt-image-1-mini, gpt-image-2), `geminiImage` (gemini-3.1-flash-image,
+gemini-3.1-flash-lite-image, gemini-3-pro-image, imagen-4.0-generate-001, etc.)
+and `byteplusImage` (Seedream — `seedream-4-0-250828`, `seedream-4-5-251128`,
+the 5.0 family).
+
+> **Use the GA Gemini image ids.** `gemini-3.1-flash-image-preview` and
+> `gemini-3-pro-image-preview` were shut down on 2026-06-25 and now 404. They
+> survive in the type union only as deprecated aliases so existing code keeps
+> compiling — a call to them typechecks and then fails at runtime. Use
+> `gemini-3.1-flash-image` / `gemini-3-pro-image` instead.
 
 > **Seedream quirks:** `watermark` defaults to **`true`** (pass
 > `modelOptions: { watermark: false }` for a clean image), `size` is a token
@@ -182,7 +189,7 @@ const openaiResult = await generateImage({
 
 // Gemini native model with aspect-ratio sizes
 const geminiResult = await generateImage({
-  adapter: geminiImage('gemini-3.1-flash-image-preview'),
+  adapter: geminiImage('gemini-3.1-flash-image'),
   prompt: 'A futuristic cityscape at night',
   size: '16:9_4K',
 })
@@ -996,7 +1003,7 @@ generateImage({
 })
 
 generateImage({
-  adapter: geminiImage('gemini-3.1-flash-image-preview'), // native multimodal
+  adapter: geminiImage('gemini-3.1-flash-image'), // native multimodal
   prompt: [
     { type: 'text', content: 'Edit this' },
     { type: 'image', source: { type: 'url', value: url } },
diff --git a/testing/e2e/global-setup.ts b/testing/e2e/global-setup.ts
index 0502e4418c..a8958ba022 100644
--- a/testing/e2e/global-setup.ts
+++ b/testing/e2e/global-setup.ts
@@ -82,6 +82,20 @@ export default async function globalSetup() {
   mock.mount('/api/embed', ollamaEmbedMount())
   mock.mount('/mistral', mistralEmbeddingsMount())
 
+  // Gemini native image generation (#1104: GA model ids + per-model
+  // aspectRatio/imageSize). Like TTS above, this hits generateContent, but
+  // aimock's handleGemini has no image-response branch at all (it imports
+  // isTextResponse/isToolCallResponse/isContentWithToolCallsResponse/
+  // isAudioResponse from helpers.js — isImageResponse is wired only into
+  // images.js's OpenAI /v1/images/* and Gemini's :predict handling) — a
+  // fixture shaped `{image}`/`{images}` falls through every branch and 500s
+  // with "Fixture response did not match any known type". Shares the
+  // '/v1beta/models' prefix with the Veo and batch-embed mounts above; only
+  // the two model paths the #1104 spec exercises are handled here, so
+  // everything else still falls through to those mounts / aimock's native
+  // Gemini handlers.
+  mock.mount('/v1beta/models', geminiNativeImageMount())
+
   // Gemini Omni Flash video generation (Interactions API). aimock handles
   // synchronous text interactions natively, but not background video jobs
   // (POST /v1beta/interactions with background:true → poll
@@ -505,6 +519,153 @@ function geminiBatchEmbedMount(): Mountable {
   }
 }
 
+/**
+ * A tiny (1x1) real PNG, base64-encoded. Only needs to be a valid inline
+ * image byte string — the #1104 spec asserts on `images.length`, not pixel
+ * content. Mirrors `FAKE_MP3_BYTES`/`FAKE_PCM_BYTES` above.
+ */
+const TINY_PNG_BASE64 =
+  'iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNk+A8AAQUBAScY42YAAAAASUVORK5CYII='
+
+/**
+ * Rejects with a Gemini-shaped error envelope (`{ error: { code, message,
+ * status } }`) so a validation failure surfaces as the actual missing/extra
+ * field in the adapter's thrown error message, not a generic 500.
+ */
+function rejectGeminiImageRequest(
+  res: http.ServerResponse,
+  message: string,
+): true {
+  res.statusCode = 400
+  res.setHeader('Content-Type', 'application/json')
+  res.end(
+    JSON.stringify({
+      error: { code: 400, message, status: 'INVALID_ARGUMENT' },
+    }),
+  )
+  return true
+}
+
+/**
+ * Mounts Gemini native image generation (`POST /v1beta/models/{model}:generateContent`)
+ * for the two models #1104's spec exercises: `gemini-3.1-flash-image` (the
+ * new GA id) and `gemini-2.5-flash-image` (the legacy model whose sizes are a
+ * bare aspect ratio with no resolution tier).
+ *
+ * aimock's own `handleGemini` has no image-response branch — it imports
+ * `isTextResponse`/`isToolCallResponse`/`isContentWithToolCallsResponse`/
+ * `isAudioResponse` from `helpers.js`, but never `isImageResponse` (that's
+ * wired only into `images.js`'s OpenAI `/v1/images/*` and Gemini's
+ * `:predict` handling) — so a `{image}`/`{images}` fixture matched against
+ * `generateContent` falls through every branch and 500s with "Fixture
+ * response did not match any known type". This mount hand-crafts the
+ * `candidates[0].content.parts[0].inlineData` shape directly, mirroring
+ * `geminiTTSMount` above (image/png instead of PCM audio).
+ *
+ * The request is validated against the raw, untranslated body — mounts see
+ * it before aimock's internal `geminiToCompletionRequest` translation, which
+ * drops `generationConfig.imageConfig` entirely — so this is what actually
+ * proves `imageConfig.aspectRatio` / `imageConfig.imageSize` reached the
+ * wire, per model:
+ * - `gemini-3.1-flash-image` rejects unless both `aspectRatio` AND
+ *   `imageSize` are present (this model has resolution tiers).
+ * - `gemini-2.5-flash-image` rejects if `aspectRatio` is missing, OR if
+ *   `imageSize` is present at all — Google documents no `image_size` for
+ *   this model (#1104), so `parseNativeImageSize` must produce a bare ratio
+ *   and the adapter must omit `imageSize` from the request.
+ */
+function geminiNativeImageMount(): Mountable {
+  const NATIVE_IMAGE_MODELS = new Set([
+    'gemini-3.1-flash-image',
+    'gemini-2.5-flash-image',
+  ])
+
+  // Deliberately coupled to the `size` values in
+  // testing/e2e/src/routes/api.gemini-image-ga-models.ts ('16:9_2K' and
+  // '16:9') — a future edit to one must update the other.
+  const EXPECTED_ASPECT_RATIO = '16:9'
+  const EXPECTED_FLASH_IMAGE_SIZE = '2K'
+
+  return {
+    async handleRequest(
+      req: http.IncomingMessage,
+      res: http.ServerResponse,
+      // aimock strips the mount prefix ('/v1beta/models'), so pathname
+      // looks like '/{model}:generateContent'.
+      pathname: string,
+    ): Promise<boolean> {
+      const match = pathname.match(/^\/([^/:]+):generateContent$/)
+      const model = match?.[1]
+      if (!model || !NATIVE_IMAGE_MODELS.has(model) || req.method !== 'POST') {
+        return false
+      }
+
+      const body = await readJsonRequestBody(req)
+      if (!body) {
+        return rejectGeminiImageRequest(res, 'Malformed JSON body.')
+      }
+
+      const generationConfig = asRecord(body.generationConfig)
+      const imageConfig = asRecord(generationConfig?.imageConfig)
+      const aspectRatio = imageConfig?.aspectRatio
+      const imageSize = imageConfig?.imageSize
+
+      if (aspectRatio !== EXPECTED_ASPECT_RATIO) {
+        return rejectGeminiImageRequest(
+          res,
+          `${model}: generationConfig.imageConfig.aspectRatio must be "${EXPECTED_ASPECT_RATIO}", got ${JSON.stringify(aspectRatio)}.`,
+        )
+      }
+
+      if (model === 'gemini-3.1-flash-image') {
+        if (imageSize !== EXPECTED_FLASH_IMAGE_SIZE) {
+          return rejectGeminiImageRequest(
+            res,
+            `${model}: generationConfig.imageConfig.imageSize must be "${EXPECTED_FLASH_IMAGE_SIZE}", got ${JSON.stringify(imageSize)}.`,
+          )
+        }
+      } else if (imageSize !== undefined) {
+        // gemini-2.5-flash-image: Google documents no image_size for this
+        // model (#1104) — the adapter must send a bare aspect ratio only.
+        return rejectGeminiImageRequest(
+          res,
+          `${model}: generationConfig.imageConfig.imageSize must be absent.`,
+        )
+      }
+
+      res.statusCode = 200
+      res.setHeader('Content-Type', 'application/json')
+      res.end(
+        JSON.stringify({
+          candidates: [
+            {
+              content: {
+                role: 'model',
+                parts: [
+                  {
+                    inlineData: {
+                      mimeType: 'image/png',
+                      data: TINY_PNG_BASE64,
+                    },
+                  },
+                ],
+              },
+              finishReason: 'STOP',
+              index: 0,
+            },
+          ],
+          usageMetadata: {
+            promptTokenCount: 8,
+            candidatesTokenCount: 1290,
+            totalTokenCount: 1298,
+          },
+        }),
+      )
+      return true
+    },
+  }
+}
+
 /**
  * Mounts Ollama's batch embed endpoint (POST /api/embed). aimock's native
  * handler answers with the legacy /api/embeddings shape (singular
diff --git a/testing/e2e/src/routeTree.gen.ts b/testing/e2e/src/routeTree.gen.ts
index c6d9853490..0cadb398d8 100644
--- a/testing/e2e/src/routeTree.gen.ts
+++ b/testing/e2e/src/routeTree.gen.ts
@@ -59,6 +59,7 @@ import { Route as ApiInterruptsTestRouteImport } from './routes/api.interrupts-t
 import { Route as ApiImageRouteImport } from './routes/api.image'
 import { Route as ApiGenerationPersistenceServerRouteImport } from './routes/api.generation-persistence-server'
 import { Route as ApiGenerationPersistenceResumeRouteImport } from './routes/api.generation-persistence-resume'
+import { Route as ApiGeminiImageGaModelsRouteImport } from './routes/api.gemini-image-ga-models'
 import { Route as ApiForeignInterruptRouteImport } from './routes/api.foreign-interrupt'
 import { Route as ApiEmbeddingRouteImport } from './routes/api.embedding'
 import { Route as ApiDurableTakeoverRouteImport } from './routes/api.durable-takeover'
@@ -336,6 +337,11 @@ const ApiGenerationPersistenceResumeRoute =
     path: '/api/generation-persistence-resume',
     getParentRoute: () => rootRouteImport,
   } as any)
+const ApiGeminiImageGaModelsRoute = ApiGeminiImageGaModelsRouteImport.update({
+  id: '/api/gemini-image-ga-models',
+  path: '/api/gemini-image-ga-models',
+  getParentRoute: () => rootRouteImport,
+} as any)
 const ApiForeignInterruptRoute = ApiForeignInterruptRouteImport.update({
   id: '/api/foreign-interrupt',
   path: '/api/foreign-interrupt',
@@ -453,6 +459,7 @@ export interface FileRoutesByFullPath {
   '/api/durable-takeover': typeof ApiDurableTakeoverRoute
   '/api/embedding': typeof ApiEmbeddingRoute
   '/api/foreign-interrupt': typeof ApiForeignInterruptRoute
+  '/api/gemini-image-ga-models': typeof ApiGeminiImageGaModelsRoute
   '/api/generation-persistence-resume': typeof ApiGenerationPersistenceResumeRoute
   '/api/generation-persistence-server': typeof ApiGenerationPersistenceServerRoute
   '/api/image': typeof ApiImageRouteWithChildren
@@ -522,6 +529,7 @@ export interface FileRoutesByTo {
   '/api/durable-takeover': typeof ApiDurableTakeoverRoute
   '/api/embedding': typeof ApiEmbeddingRoute
   '/api/foreign-interrupt': typeof ApiForeignInterruptRoute
+  '/api/gemini-image-ga-models': typeof ApiGeminiImageGaModelsRoute
   '/api/generation-persistence-resume': typeof ApiGenerationPersistenceResumeRoute
   '/api/generation-persistence-server': typeof ApiGenerationPersistenceServerRoute
   '/api/image': typeof ApiImageRouteWithChildren
@@ -592,6 +600,7 @@ export interface FileRoutesById {
   '/api/durable-takeover': typeof ApiDurableTakeoverRoute
   '/api/embedding': typeof ApiEmbeddingRoute
   '/api/foreign-interrupt': typeof ApiForeignInterruptRoute
+  '/api/gemini-image-ga-models': typeof ApiGeminiImageGaModelsRoute
   '/api/generation-persistence-resume': typeof ApiGenerationPersistenceResumeRoute
   '/api/generation-persistence-server': typeof ApiGenerationPersistenceServerRoute
   '/api/image': typeof ApiImageRouteWithChildren
@@ -663,6 +672,7 @@ export interface FileRouteTypes {
     | '/api/durable-takeover'
     | '/api/embedding'
     | '/api/foreign-interrupt'
+    | '/api/gemini-image-ga-models'
     | '/api/generation-persistence-resume'
     | '/api/generation-persistence-server'
     | '/api/image'
@@ -732,6 +742,7 @@ export interface FileRouteTypes {
     | '/api/durable-takeover'
     | '/api/embedding'
     | '/api/foreign-interrupt'
+    | '/api/gemini-image-ga-models'
     | '/api/generation-persistence-resume'
     | '/api/generation-persistence-server'
     | '/api/image'
@@ -801,6 +812,7 @@ export interface FileRouteTypes {
     | '/api/durable-takeover'
     | '/api/embedding'
     | '/api/foreign-interrupt'
+    | '/api/gemini-image-ga-models'
     | '/api/generation-persistence-resume'
     | '/api/generation-persistence-server'
     | '/api/image'
@@ -871,6 +883,7 @@ export interface RootRouteChildren {
   ApiDurableTakeoverRoute: typeof ApiDurableTakeoverRoute
   ApiEmbeddingRoute: typeof ApiEmbeddingRoute
   ApiForeignInterruptRoute: typeof ApiForeignInterruptRoute
+  ApiGeminiImageGaModelsRoute: typeof ApiGeminiImageGaModelsRoute
   ApiGenerationPersistenceResumeRoute: typeof ApiGenerationPersistenceResumeRoute
   ApiGenerationPersistenceServerRoute: typeof ApiGenerationPersistenceServerRoute
   ApiImageRoute: typeof ApiImageRouteWithChildren
@@ -1258,6 +1271,13 @@ declare module '@tanstack/react-router' {
       preLoaderRoute: typeof ApiGenerationPersistenceResumeRouteImport
       parentRoute: typeof rootRouteImport
     }
+    '/api/gemini-image-ga-models': {
+      id: '/api/gemini-image-ga-models'
+      path: '/api/gemini-image-ga-models'
+      fullPath: '/api/gemini-image-ga-models'
+      preLoaderRoute: typeof ApiGeminiImageGaModelsRouteImport
+      parentRoute: typeof rootRouteImport
+    }
     '/api/foreign-interrupt': {
       id: '/api/foreign-interrupt'
       path: '/api/foreign-interrupt'
@@ -1468,6 +1488,7 @@ const rootRouteChildren: RootRouteChildren = {
   ApiDurableTakeoverRoute: ApiDurableTakeoverRoute,
   ApiEmbeddingRoute: ApiEmbeddingRoute,
   ApiForeignInterruptRoute: ApiForeignInterruptRoute,
+  ApiGeminiImageGaModelsRoute: ApiGeminiImageGaModelsRoute,
   ApiGenerationPersistenceResumeRoute: ApiGenerationPersistenceResumeRoute,
   ApiGenerationPersistenceServerRoute: ApiGenerationPersistenceServerRoute,
   ApiImageRoute: ApiImageRouteWithChildren,
diff --git a/testing/e2e/src/routes/api.gemini-image-ga-models.ts b/testing/e2e/src/routes/api.gemini-image-ga-models.ts
new file mode 100644
index 0000000000..645497eb57
--- /dev/null
+++ b/testing/e2e/src/routes/api.gemini-image-ga-models.ts
@@ -0,0 +1,98 @@
+import { createFileRoute } from '@tanstack/react-router'
+import { generateImage } from '@tanstack/ai'
+import { createGeminiImage, GeminiImageModels } from '@tanstack/ai-gemini'
+import type { ImageGenerationResult } from '@tanstack/ai'
+
+const LLMOCK_DEFAULT_BASE = process.env.LLMOCK_URL || 'http://127.0.0.1:4010'
+const DUMMY_KEY = 'sk-e2e-test-dummy-key'
+
+interface StageResult {
+  imageCount?: number
+  error?: string
+}
+
+async function runStage(
+  generate: () => Promise<ImageGenerationResult>,
+): Promise<StageResult> {
+  try {
+    const result = await generate()
+    return { imageCount: result.images.length }
+  } catch (error) {
+    return { error: error instanceof Error ? error.message : String(error) }
+  }
+}
+
+/**
+ * Regression coverage for #1104 (GA image model ids + per-model sizes).
+ *
+ * Runs the two model calls as independent stages (each with its own
+ * try/catch) so a regression in one doesn't mask the other — the two specs
+ * each assert on only their own stage's result.
+ *
+ * 1. `gaModel`: `gemini-3.1-flash-image` — the GA id for "Nano Banana 2" —
+ *    must be listed in the package's exported `GeminiImageModels` runtime
+ *    array and must actually route a `generateImage()` call through
+ *    Gemini's `generateContent` API to a returned image. Before the fix,
+ *    only the shut-down `-preview` id was in that list (the GA id 404s
+ *    against the real API otherwise). `GeminiImageModels` is genuine
+ *    runtime data (an exported `as const` array), unlike the per-model
+ *    `size` typing below, so `.includes()` on it is real, revert-detectable
+ *    coverage independent of any TypeScript compile step — nothing in the
+ *    adapter's routing logic gates on this array at runtime, so it's the
+ *    only way to observe the id addition without a type-checker.
+ * 2. `legacyModel`: `gemini-2.5-flash-image` must send
+ *    `generationConfig.imageConfig` with only `aspectRatio` set — no
+ *    `imageSize` — because Google documents no `image_size` value for this
+ *    model. `parseNativeImageSize`'s regex used to require an
+ *    `_<resolution>` suffix on every size string, so a bare ratio like
+ *    `"16:9"` failed to match at all pre-fix and no `imageConfig` (not even
+ *    `aspectRatio`) reached the wire. The `geminiNativeImageMount` in
+ *    `global-setup.ts` 400s unless `aspectRatio` is present and `imageSize`
+ *    is absent, so this is genuine runtime proof of the parser change, not
+ *    just a type check.
+ *
+ * Both models are proxied through `geminiNativeImageMount` (global-setup.ts)
+ * — aimock's own Gemini `generateContent` handler has no image-response
+ * branch (see that mount's doc comment for why).
+ */
+export const Route = createFileRoute('/api/gemini-image-ga-models')({
+  server: {
+    handlers: {
+      POST: async () => {
+        const gaModelListed = GeminiImageModels.includes(
+          'gemini-3.1-flash-image',
+        )
+
+        const gaModel = await runStage(() =>
+          generateImage({
+            adapter: createGeminiImage('gemini-3.1-flash-image', DUMMY_KEY, {
+              httpOptions: { baseUrl: LLMOCK_DEFAULT_BASE },
+            }),
+            prompt: 'a guitar in a music store',
+            size: '16:9_2K',
+          }),
+        )
+
+        const legacyModel = await runStage(() =>
+          generateImage({
+            adapter: createGeminiImage('gemini-2.5-flash-image', DUMMY_KEY, {
+              httpOptions: { baseUrl: LLMOCK_DEFAULT_BASE },
+            }),
+            prompt: 'a guitar in a music store',
+            // Bare aspect ratio, no `_<resolution>` suffix — the shape
+            // `gemini-2.5-flash-image` requires post-#1104.
+            size: '16:9',
+          }),
+        )
+
+        return new Response(
+          JSON.stringify({ gaModelListed, gaModel, legacyModel }),
+          {
+            status: 200,
+            headers: { 'Content-Type': 'application/json' },
+          },
+        )
+      },
+    },
+  },
+})
diff --git a/testing/e2e/tests/gemini-image-ga-models.spec.ts b/testing/e2e/tests/gemini-image-ga-models.spec.ts
new file mode 100644
index 0000000000..8219386b43
--- /dev/null
+++ b/testing/e2e/tests/gemini-image-ga-models.spec.ts
@@ -0,0 +1,56 @@
+import { test, expect } from './fixtures'
+
+interface StageResult {
+  imageCount?: number
+  error?: string
+}
+
+interface GeminiImageGaModelsResponse {
+  gaModelListed: boolean
+  gaModel: StageResult
+  legacyModel: StageResult
+}
+
+/**
+ * Regression coverage for #1104 (Gemini GA image model ids + per-model
+ * sizes). See `api.gemini-image-ga-models.ts` for the full mechanism and
+ * `geminiNativeImageMount` in `global-setup.ts` for the mock that makes it
+ * possible — aimock's native Gemini `generateContent` handler has no
+ * image-response branch, so this route/mount pair exists specifically to
+ * cover what aimock alone cannot. The route runs both model calls as
+ * independent stages, so each test below only depends on its own stage.
+ */
+test.describe('gemini — GA image model ids and per-model sizes (#1104)', () => {
+  test('gemini-3.1-flash-image is a listed GA id and generates an image', async ({
+    request,
+  }) => {
+    const res = await request.post('/api/gemini-image-ga-models')
+    expect(res.ok()).toBe(true)
+
+    const { gaModelListed, gaModel } =
+      (await res.json()) as GeminiImageGaModelsResponse
+
+    // Genuine runtime data, not a type check: `GeminiImageModels` is the
+    // package's exported `as const` array of supported image model ids.
+    // Nothing at runtime gates request routing on this array (the adapter
+    // dispatches on the model string's "gemini-" prefix, not membership), so
+    // checking its contents is the only way to observe the id addition
+    // without a TypeScript compile step. Before #1104 it only carried the
+    // shut-down `gemini-3.1-flash-image-preview` alias.
+    expect(gaModelListed).toBe(true)
+    expect(gaModel.error ?? null).toBeNull()
+    expect(gaModel.imageCount).toBe(1)
+  })
+
+  test('gemini-2.5-flash-image sends imageConfig.aspectRatio with no imageSize', async ({
+    request,
+  }) => {
+    const res = await request.post('/api/gemini-image-ga-models')
+    expect(res.ok()).toBe(true)
+
+    const { legacyModel } = (await res.json()) as GeminiImageGaModelsResponse
+
+    expect(legacyModel.error ?? null).toBeNull()
+    expect(legacyModel.imageCount).toBe(1)
+  })
+})
diff --git a/testing/panel/src/routes/api.image.ts b/testing/panel/src/routes/api.image.ts
index e9c4aeda48..1a71332e3f 100644
--- a/testing/panel/src/routes/api.image.ts
+++ b/testing/panel/src/routes/api.image.ts
@@ -19,7 +19,7 @@ export const Route = createFileRoute('/api/image')({
         const defaultModels: Record<Provider, string> = {
           openai: 'gpt-image-1',
           gemini: 'gemini-2.5-flash-image',
-          openrouter: 'google/gemini-3.1-flash-image-preview',
+          openrouter: 'google/gemini-3.1-flash-image',
         }
         const model: string =
           data.model || body.model || defaultModels[provider]