diff --git a/src/jarvis/reply/engine.py b/src/jarvis/reply/engine.py
index d04c41c..39f4620 100644
--- a/src/jarvis/reply/engine.py
+++ b/src/jarvis/reply/engine.py
@@ -889,6 +889,37 @@ def _maybe_deterministic_site_search(text: str, db, cfg, language) -> Optional[s
     return None
 
 
+_WEATHER_INTENT_WORDS = (
+    "날씨", "기온", "더워", "더운", "추워", "추운", "비 와", "비와", "비 올",
+    "눈 와", "눈와", "weather", "temperature", "forecast",
+)
+
+
+def _maybe_deterministic_weather(text: str, db, cfg, language) -> Optional[str]:
+    """Run getWeather directly and return its concise Korean sentence, bypassing
+    the chat model. The 7B otherwise re-synthesises the weather into multiple
+    sentences and leaks units ("25도 Celsius"); the tool already formats one
+    clean Korean sentence, so for a plain weather ask we just return it.
+    Fail-open: any problem returns None and the normal flow continues.
+    """
+    try:
+        low = (text or "").lower()
+        if not any(w in low for w in _WEATHER_INTENT_WORDS):
+            return None
+        from ..tools.registry import run_tool_with_retries
+        res = run_tool_with_retries(
+            db=db, cfg=cfg, tool_name="getWeather", tool_args={},
+            system_prompt="", original_prompt="", redacted_text=redact(text),
+            max_retries=1, language=language,
+        )
+        if res and getattr(res, "success", False) and res.reply_text:
+            debug_log("deterministic weather executed", "tools")
+            return res.reply_text
+    except Exception as e:  # noqa: BLE001
+        debug_log(f"deterministic weather failed (fail-open): {e}", "tools")
+    return None
+
+
 def run_reply_engine(db: "Database", cfg, tts: Optional[Any],
                     text: str, dialogue_memory: "DialogueMemory",
                     language: Optional[str] = None) -> Optional[str]:
@@ -920,6 +951,13 @@ def run_reply_engine(db: "Database", cfg, tts: Optional[Any],
     if _site_search_reply is not None:
         return _site_search_reply
 
+    # Step 0.6: Deterministic weather — return getWeather's concise Korean
+    # sentence directly so the chat model can't rephrase it into multiple
+    # sentences or leak units.
+    _weather_reply = _maybe_deterministic_weather(text, db, cfg, language)
+    if _weather_reply is not None:
+        return _weather_reply
+
     # Step 2: Check for recent dialogue context
     recent_messages = []
     is_new_conversation = True
diff --git a/src/jarvis/tools/builtin/weather.py b/src/jarvis/tools/builtin/weather.py
index dab8694..36b495f 100644
--- a/src/jarvis/tools/builtin/weather.py
+++ b/src/jarvis/tools/builtin/weather.py
@@ -440,17 +440,11 @@ class WeatherTool(Tool):
                 _ko_parts.append(_t)
             ko_sentence = ", ".join(_ko_parts) + "입니다."
 
-            # Build response text — concise current conditions (Korean sentence
-            # first so the model echoes it; English detail kept for any follow-up
-            # reasoning but the forecast firehose is dropped to curb rambling).
-            lines = [
-                f"한국어로 정확히 이 한 문장만 답하세요: {ko_sentence}",
-                f"(참고 데이터 — 답변에 추가하지 말 것: {location_display}, {weather_desc}, "
-                f"{temp_c}°C feels {feels_like_c}°C, humidity {humidity}%, wind {wind_speed}km/h)",
-            ]
-            # Forecast (hourly / 7-day) is intentionally omitted from the default
-            # voice reply to keep it to one spoken sentence; current conditions
-            # are what "날씨 알려줘" asks for.
+            # The reply is the clean Korean sentence ONLY — no English/°C source
+            # for the model to echo ("25도 Celsius"), no forecast firehose to
+            # ramble over. The deterministic weather path in the engine returns
+            # this verbatim; on the LLM path the model just echoes one sentence.
+            lines = [ko_sentence]
 
             reply_text = "\n".join(lines)