diff --git a/backend/routers/voice.py b/backend/routers/voice.py
index 72a7ceb..50c7e2c 100644
--- a/backend/routers/voice.py
+++ b/backend/routers/voice.py
@@ -208,6 +208,10 @@ async def voice_chat(body: ChatIn) -> StreamingResponse:
             user_text = f"[Bildschirm-Sicht — das ist gerade auf dem/den Schirm(en) zu sehen:\n{safe_desc}\n]\n\n{body.text}"
     messages.append({"role": "user", "content": user_text})
     payload = {"model": body.model or HERMES_API_MODEL, "messages": messages, "stream": True}
+    # Lucy-Hirn ist Thinking-Modell -> fuer die gesprochene Assistentin Thinking AUS (sonst 11k
+    # Reasoning-Token vor der kurzen Antwort, gemessen ~30s TTFB). Muster wie gateway_proxy/mem0.
+    if os.environ.get("MC_VOICE_NO_THINK", "1") not in ("0", "false", "False"):
+        payload["chat_template_kwargs"] = {"enable_thinking": False}
     headers = {
         "Authorization": f"Bearer {HERMES_API_KEY}",
         "X-Hermes-Session-Id": body.session_id,