diff --git a/extensions/smart-router/index.ts b/extensions/smart-router/index.ts
index 1747c74..5590508 100644
--- a/extensions/smart-router/index.ts
+++ b/extensions/smart-router/index.ts
@@ -30,24 +30,34 @@ const COMPRESS_MIN_CHARS = 20_000;
 let currentTag: string | null = null;
 
 // Model ID mappings for routing
-// Cost-conscious: flash/economy for most tasks, Pro only for genuinely complex work
+// TEMPORARY: All routes set to owl-alpha to stop cost drain.
+// Restore cost-aware routing by replacing owl-alpha with the models commented below.
+//
+// RESTORE MAP (uncomment and replace owl-alpha lines):
+//   "free-core":          { provider: "openrouter", id: "free" },                    // was: openrouter/free (keep)
+//   "router-eval":        { provider: "openrouter", id: "free" },                    // was: openrouter/free (keep)
+//   "economy-devops":     { provider: "opencode-go", id: "deepseek/deepseek-v4-flash" }, // cheap
+//   "economy-code":       { provider: "opencode-go", id: "deepseek/deepseek-v4-flash" }, // cheap
+//   "precision-devops":   { provider: "opencode-go", id: "deepseek/deepseek-v4-flash" }, // cheap
+//   "precision-code-high":{ provider: "opencode-go", id: "deepseek/deepseek-v4-pro" },   // expensive, only for complex code
+//   "precision-react":    { provider: "opencode-go", id: "deepseek/deepseek-v4-pro" },   // expensive, only for complex React
+//   "context-heavy":      { provider: "openrouter", id: "free" },                    // was: openrouter/free (keep)
+//
+// THINKING restore: precision-code-high and precision-react → "high", everything else → off
 const MODELS: Record<string, { provider: string; id: string }> = {
-  "free-core":          { provider: "openrouter", id: "free" },
-  "router-eval":        { provider: "openrouter", id: "free" },
-  "economy-devops":     { provider: "opencode-go", id: "deepseek/deepseek-v4-flash" },
-  "economy-code":       { provider: "opencode-go", id: "deepseek/deepseek-v4-flash" },
-  "precision-devops":   { provider: "opencode-go", id: "deepseek/deepseek-v4-flash" },
-  "precision-code-high":{ provider: "opencode-go", id: "deepseek/deepseek-v4-pro" },
-  "precision-react":    { provider: "opencode-go", id: "deepseek/deepseek-v4-pro" },
-  "context-heavy":      { provider: "openrouter", id: "free" },
+  "free-core":          { provider: "openrouter", id: "owl-alpha" },
+  "router-eval":        { provider: "openrouter", id: "owl-alpha" },
+  "economy-devops":     { provider: "openrouter", id: "owl-alpha" },
+  "economy-code":       { provider: "openrouter", id: "owl-alpha" },
+  "precision-devops":   { provider: "openrouter", id: "owl-alpha" },
+  "precision-code-high":{ provider: "openrouter", id: "owl-alpha" },
+  "precision-react":    { provider: "openrouter", id: "owl-alpha" },
+  "context-heavy":      { provider: "openrouter", id: "owl-alpha" },
 };
 
-// Thinking level mappings — only set for tasks that truly benefit.
-// Everything else keeps the default (off) for speed and cost.
-const THINKING: Record<string, string> = {
-  "precision-code-high":"high",
-  "precision-react":    "high",
-};
+// TEMPORARY: All thinking disabled while using owl-alpha.
+// Restore: precision-code-high and precision-react → "high", everything else → off.
+const THINKING: Record<string, string> = {};
 
 function getModel(ctx: ExtensionContext, key: string): any {
   const mapping = MODELS[key];