diff --git a/packages/opencode/src/provider/provider.ts b/packages/opencode/src/provider/provider.ts
index 8f6e1556ada3..6267706c3364 100644
--- a/packages/opencode/src/provider/provider.ts
+++ b/packages/opencode/src/provider/provider.ts
@@ -1130,16 +1130,27 @@ const layer: Layer.Layer<
               if (model.id && model.id !== modelID) return modelID
               return existingModel?.name ?? modelID
             })
+            const resolvedNpm =
+              model.provider?.npm ??
+              provider.npm ??
+              existingModel?.api.npm ??
+              modelsDev[providerID]?.npm ??
+              "@ai-sdk/openai-compatible"
+            // For custom models on OpenAI-compatible providers, default image input
+            // to true since the SDK natively converts file parts to image_url format.
+            // This ensures custom provider models support image attachments out of the
+            // box without requiring explicit modalities configuration.
+            const isOpenAICompatible =
+              resolvedNpm === "@ai-sdk/openai-compatible" ||
+              resolvedNpm === "@ai-sdk/openai" ||
+              resolvedNpm === "@ai-sdk/azure" ||
+              resolvedNpm === "@ai-sdk/github-copilot"
+            const defaultImageInput = isOpenAICompatible ? true : false
             const parsedModel: Model = {
               id: ModelID.make(modelID),
               api: {
                 id: model.id ?? existingModel?.api.id ?? modelID,
-                npm:
-                  model.provider?.npm ??
-                  provider.npm ??
-                  existingModel?.api.npm ??
-                  modelsDev[providerID]?.npm ??
-                  "@ai-sdk/openai-compatible",
+                npm: resolvedNpm,
                 url: model.provider?.api ?? provider?.api ?? existingModel?.api.url ?? modelsDev[providerID]?.api ?? "",
               },
               status: model.status ?? existingModel?.status ?? "active",
@@ -1148,12 +1159,15 @@ const layer: Layer.Layer<
               capabilities: {
                 temperature: model.temperature ?? existingModel?.capabilities.temperature ?? false,
                 reasoning: model.reasoning ?? existingModel?.capabilities.reasoning ?? false,
-                attachment: model.attachment ?? existingModel?.capabilities.attachment ?? false,
+                attachment: model.attachment ?? existingModel?.capabilities.attachment ?? defaultImageInput,
                 toolcall: model.tool_call ?? existingModel?.capabilities.toolcall ?? true,
                 input: {
                   text: model.modalities?.input?.includes("text") ?? existingModel?.capabilities.input.text ?? true,
                   audio: model.modalities?.input?.includes("audio") ?? existingModel?.capabilities.input.audio ?? false,
-                  image: model.modalities?.input?.includes("image") ?? existingModel?.capabilities.input.image ?? false,
+                  image:
+                    model.modalities?.input?.includes("image") ??
+                    existingModel?.capabilities.input.image ??
+                    defaultImageInput,
                   video: model.modalities?.input?.includes("video") ?? existingModel?.capabilities.input.video ?? false,
                   pdf: model.modalities?.input?.includes("pdf") ?? existingModel?.capabilities.input.pdf ?? false,
                 },
diff --git a/packages/opencode/test/provider/provider.test.ts b/packages/opencode/test/provider/provider.test.ts
index 8993020820e3..de17c00682e6 100644
--- a/packages/opencode/test/provider/provider.test.ts
+++ b/packages/opencode/test/provider/provider.test.ts
@@ -922,6 +922,55 @@ test("model modalities default correctly", async () => {
       const model = providers[ProviderID.make("test-provider")].models["test-model"]
       expect(model.capabilities.input.text).toBe(true)
       expect(model.capabilities.output.text).toBe(true)
+      // OpenAI-compatible providers should default image input to true
+      expect(model.capabilities.input.image).toBe(true)
+      expect(model.capabilities.attachment).toBe(true)
+    },
+  })
+})
+
+test("custom openai-compatible provider defaults image input to true", async () => {
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      await Bun.write(
+        path.join(dir, "opencode.json"),
+        JSON.stringify({
+          $schema: "https://opencode.ai/config.json",
+          provider: {
+            "custom-vision": {
+              name: "Custom Vision Provider",
+              npm: "@ai-sdk/openai-compatible",
+              env: [],
+              models: {
+                "vision-model": {
+                  name: "Vision Model",
+                  tool_call: true,
+                  limit: { context: 128000, output: 4096 },
+                  // No modalities specified - should default image to true
+                },
+              },
+              options: {
+                apiKey: "test-key",
+                baseURL: "https://api.custom.com/v1",
+              },
+            },
+          },
+        }),
+      )
+    },
+  })
+  await Instance.provide({
+    directory: tmp.path,
+    fn: async () => {
+      const providers = await list()
+      const model = providers[ProviderID.make("custom-vision")].models["vision-model"]
+      // Should default to true for OpenAI-compatible providers
+      expect(model.capabilities.input.image).toBe(true)
+      expect(model.capabilities.attachment).toBe(true)
+      // Other modalities should still default to false
+      expect(model.capabilities.input.audio).toBe(false)
+      expect(model.capabilities.input.video).toBe(false)
+      expect(model.capabilities.input.pdf).toBe(false)
     },
   })
 })