diff --git a/src/LLMProviders/embeddingManager.ts b/src/LLMProviders/embeddingManager.ts
index c46030925..453133559 100644
--- a/src/LLMProviders/embeddingManager.ts
+++ b/src/LLMProviders/embeddingManager.ts
@@ -9,11 +9,12 @@ import { CohereEmbeddings } from "@langchain/cohere";
 import { Embeddings } from "@langchain/core/embeddings";
 import { GoogleGenerativeAIEmbeddings } from "@langchain/google-genai";
 import { OllamaEmbeddings } from "@langchain/ollama";
-import { AzureOpenAIEmbeddings, OpenAIEmbeddings } from "@langchain/openai";
+import { OpenAIEmbeddings } from "@langchain/openai";
 import { Notice } from "obsidian";
 import { BrevilabsClient } from "./brevilabsClient";
 import { CustomJinaEmbeddings } from "./CustomJinaEmbeddings";
 import { CustomOpenAIEmbeddings } from "./CustomOpenAIEmbeddings";
+import { normalizeAzureUrl } from "./chatModelManager";
 
 type EmbeddingConstructorType = new (config: any) => Embeddings;
 
@@ -23,7 +24,7 @@ const EMBEDDING_PROVIDER_CONSTRUCTORS = {
   [EmbeddingModelProviders.OPENAI]: OpenAIEmbeddings,
   [EmbeddingModelProviders.COHEREAI]: CohereEmbeddings,
   [EmbeddingModelProviders.GOOGLE]: GoogleGenerativeAIEmbeddings,
-  [EmbeddingModelProviders.AZURE_OPENAI]: AzureOpenAIEmbeddings,
+  [EmbeddingModelProviders.AZURE_OPENAI]: OpenAIEmbeddings,
   [EmbeddingModelProviders.OLLAMA]: OllamaEmbeddings,
   [EmbeddingModelProviders.LM_STUDIO]: CustomOpenAIEmbeddings,
   [EmbeddingModelProviders.OPENAI_FORMAT]: OpenAIEmbeddings,
@@ -248,16 +249,60 @@ export default class EmbeddingManager {
         modelName: modelName,
         apiKey: await getDecryptedKey(settings.googleApiKey),
       },
-      [EmbeddingModelProviders.AZURE_OPENAI]: {
-        modelName,
-        azureOpenAIApiKey: await getDecryptedKey(customModel.apiKey || settings.azureOpenAIApiKey),
-        azureOpenAIApiInstanceName:
-          customModel.azureOpenAIApiInstanceName || settings.azureOpenAIApiInstanceName,
-        azureOpenAIApiDeploymentName:
+      [EmbeddingModelProviders.AZURE_OPENAI]: await (async () => {
+        const decryptedApiKey = await getDecryptedKey(
+          customModel.apiKey || settings.azureOpenAIApiKey
+        );
+
+        const hasCustomBaseUrl = Boolean(customModel.baseUrl?.trim());
+        const apiVersion = customModel.azureOpenAIApiVersion || settings.azureOpenAIApiVersion;
+
+        if (hasCustomBaseUrl) {
+          const azureUrl = normalizeAzureUrl(customModel.baseUrl!.trim());
+
+          return {
+            modelName,
+            apiKey: decryptedApiKey,
+            timeout: 10000,
+            batchSize: getSettings().embeddingBatchSize,
+            configuration: {
+              baseURL: azureUrl.baseUrl || customModel.baseUrl!.trim(),
+              defaultQuery: {
+                "api-version": azureUrl.apiVersion || apiVersion,
+              },
+              fetch: customModel.enableCors ? safeFetch : undefined,
+            },
+          } as ExtendedConfig<
+            ConstructorParameters<
+              EmbeddingProviderConstructorMap[typeof EmbeddingModelProviders.AZURE_OPENAI]
+            >[0]
+          >;
+        }
+
+        const instanceName =
+          customModel.azureOpenAIApiInstanceName || settings.azureOpenAIApiInstanceName;
+        const deploymentName =
           customModel.azureOpenAIApiEmbeddingDeploymentName ||
-          settings.azureOpenAIApiEmbeddingDeploymentName,
-        azureOpenAIApiVersion: customModel.azureOpenAIApiVersion || settings.azureOpenAIApiVersion,
-      },
+          settings.azureOpenAIApiEmbeddingDeploymentName;
+
+        return {
+          modelName,
+          apiKey: decryptedApiKey,
+          timeout: 10000,
+          batchSize: getSettings().embeddingBatchSize,
+          configuration: {
+            baseURL: `https://${instanceName}.openai.azure.com/openai/deployments/${deploymentName}`,
+            defaultQuery: {
+              "api-version": apiVersion,
+            },
+            fetch: customModel.enableCors ? safeFetch : undefined,
+          },
+        } as ExtendedConfig<
+          ConstructorParameters<
+            EmbeddingProviderConstructorMap[typeof EmbeddingModelProviders.AZURE_OPENAI]
+          >[0]
+        >;
+      })(),
       [EmbeddingModelProviders.OLLAMA]: {
         baseUrl: customModel.baseUrl || "http://localhost:11434",
         model: modelName,
diff --git a/src/settings/v2/components/ModelAddDialog.tsx b/src/settings/v2/components/ModelAddDialog.tsx
index 40d49065b..254043ed5 100644
--- a/src/settings/v2/components/ModelAddDialog.tsx
+++ b/src/settings/v2/components/ModelAddDialog.tsx
@@ -115,12 +115,12 @@ export const ModelAddDialog: React.FC<ModelAddDialogProps> = ({
     if (!model.name) isValid = false;
 
     // Validate Azure OpenAI specific fields.
-    // Embedding models always require the legacy fields because EmbeddingManager
-    // does not consume baseUrl and still reads azureOpenAIApiInstanceName,
-    // azureOpenAIApiEmbeddingDeploymentName, and azureOpenAIApiVersion directly.
-    // Chat models may skip legacy fields when a full base URL is supplied instead.
+    // When a Base URL is provided, both chat and embedding models can
+    // skip the legacy Azure fields and rely on the normalized endpoint.
+    // When Base URL is empty, fall back to legacy instance/deployment/version
+    // fields for backward compatibility.
     const isAzure = model.provider === ChatModelProviders.AZURE_OPENAI;
-    const azureRequiresLegacyFields = isAzure && (isEmbeddingModel || !model.baseUrl?.trim());
+    const azureRequiresLegacyFields = isAzure && !model.baseUrl?.trim();
     if (azureRequiresLegacyFields) {
       newErrors.instanceName = !model.azureOpenAIApiInstanceName;
       newErrors.apiVersion = !model.azureOpenAIApiVersion;
@@ -340,10 +340,10 @@ export const ModelAddDialog: React.FC<ModelAddDialogProps> = ({
             </FormField>
           );
         case ChatModelProviders.AZURE_OPENAI:
-          // Chat models with a base URL use the new flow and skip legacy fields.
-          // Embedding models always require legacy fields since EmbeddingManager
-          // reads them directly and does not consume baseUrl.
-          if (model.baseUrl?.trim() && !isEmbeddingModel) return null;
+          // Chat and embedding models with a Base URL use the new flow
+          // and skip legacy fields. When Base URL is empty, fall back to
+          // legacy instance/deployment/version inputs.
+          if (model.baseUrl?.trim()) return null;
           return (
             <>
               <FormField
@@ -500,14 +500,11 @@ export const ModelAddDialog: React.FC<ModelAddDialogProps> = ({
       return providerInfo.host;
     }
 
-    const instanceName = model.azureOpenAIApiInstanceName || "[instance]";
-    const deploymentName = isEmbeddingModel
-      ? model.azureOpenAIApiEmbeddingDeploymentName || "[deployment]"
-      : model.azureOpenAIApiDeploymentName || "[deployment]";
-    const apiVersion = model.azureOpenAIApiVersion || "[api-version]";
-    const endpoint = isEmbeddingModel ? "embeddings" : "chat/completions";
-
-    return `https://${instanceName}.openai.azure.com/openai/deployments/${deploymentName}/${endpoint}?api-version=${apiVersion}`;
+    // For Azure, encourage users to paste either the full endpoint URL
+    // (including /embeddings or /chat/completions and api-version) or
+    // the deployment base URL. Legacy instance/deployment/version fields
+    // are still available below when Base URL is empty.
+    return "https://[proxy]/deployments/[model]/embeddings?api-version=[version]";
   };
 
   const capabilityOptions = Object.entries(MODEL_CAPABILITIES).map(([id, description]) => ({