diff --git a/体己/AI/AIRuntime.swift b/体己/AI/AIRuntime.swift
index 4efbcee..cdc518a 100644
--- a/体己/AI/AIRuntime.swift
+++ b/体己/AI/AIRuntime.swift
@@ -34,9 +34,15 @@ actor AIRuntime {
     /// 加载模型。首次调用会真正加载,后续幂等。
     func prepare() async throws {
         switch status {
-        case .ready: return
-        case .loading: return  // 已经在加载
-        case .error, .notReady: break
+        case .ready:
+            return
+        case .loading:
+            // 已有其他调用方在加载;本次 prepare 直接返回,
+            // 调用方需稍后 await prepare() 再判 status,或自行轮询 / 显示加载 UI。
+            // W3 引入 prepare 队列时优化。
+            return
+        case .error, .notReady:
+            break
         }
 
         guard ModelStore.shared.isReady(.llm) else {
@@ -65,7 +71,7 @@ actor AIRuntime {
         let snapshotSession = llmSession
 
         return AsyncThrowingStream { continuation in
-            Task { [weak self] in
+            Task {
                 guard snapshotStatus == .ready, let session = snapshotSession else {
                     continuation.finish(throwing: AIRuntimeError.notReady)
                     return
@@ -74,7 +80,7 @@ actor AIRuntime {
                     // session.generate 跨 actor 边界,需要 await
                     let stream = await session.generate(prompt: prompt, maxTokens: maxTokens)
                     for try await chunk in stream {
-                        await self?.recordRate(chunk.decodeRate)
+                        await self.recordRate(chunk.decodeRate)
                         continuation.yield(chunk)
                     }
                     continuation.finish()