diff --git a/apps/api/src/hermes/nl_gateway.py b/apps/api/src/hermes/nl_gateway.py
index e68d9b53..b665a75a 100644
--- a/apps/api/src/hermes/nl_gateway.py
+++ b/apps/api/src/hermes/nl_gateway.py
@@ -54,21 +54,22 @@ _RATE_LIMIT_WINDOW_SEC = 60
 # Ollama 模型路由（按 agent 專業選最適模型，111 主機）
 # ─────────────────────────────────────────────────────────────────────────────
 _MODEL_BY_AGENT: dict[str, str] = {
-    # 推理型（找根因 / 安全分析）→ deepseek-r1:14b
+    # 推理型（找根因 / 安全分析）→ deepseek-r1:14b（CoT 推理）
     "debugger":           "deepseek-r1:14b",
     "vuln-verifier":      "deepseek-r1:14b",
-    # 程式碼專用（review / 實作 / 重構 / DB / 前端 / 工具）→ qwen2.5-coder:7b
-    "critic":             "qwen2.5-coder:7b",
-    "db-expert":          "qwen2.5-coder:7b",
-    "fullstack-engineer": "qwen2.5-coder:7b",
-    "refactor-specialist":"qwen2.5-coder:7b",
-    "migration-engineer": "qwen2.5-coder:7b",
-    "frontend-designer":  "qwen2.5-coder:7b",
-    "tool-expert":        "qwen2.5-coder:7b",
-    # 通用指令（規劃 / 導覽 / 文件）→ qwen2.5:7b-instruct
-    "planner":            "qwen2.5:7b-instruct",
-    "onboarder":          "qwen2.5:7b-instruct",
-    "web-researcher":     "qwen2.5:7b-instruct",
+    # 程式碼 + 通用（review / 實作 / 重構 / DB / 前端 / 工具 / 規劃 / 文件）→ qwen3:8b
+    # 2026-04-25 ogt + Claude Sonnet 4.6: qwen2.5-coder:7b + qwen2.5:7b-instruct → qwen3:8b
+    #   qwen3:8b Hybrid Thinking 同時勝任程式碼與指令；gemma4 尚未在 Ollama 釋出
+    "critic":             "qwen3:8b",
+    "db-expert":          "qwen3:8b",
+    "fullstack-engineer": "qwen3:8b",
+    "refactor-specialist":"qwen3:8b",
+    "migration-engineer": "qwen3:8b",
+    "frontend-designer":  "qwen3:8b",
+    "tool-expert":        "qwen3:8b",
+    "planner":            "qwen3:8b",
+    "onboarder":          "qwen3:8b",
+    "web-researcher":     "qwen3:8b",
 }
 _DEFAULT_MODEL = "deepseek-r1:14b"
 _OLLAMA_TIMEOUT = 90.0  # deepseek-r1:14b 推理較慢，給 90s