diff --git a/backend/CLAUDE.md b/backend/CLAUDE.md
index caa36f579..8490d8644 100644
--- a/backend/CLAUDE.md
+++ b/backend/CLAUDE.md
@@ -263,7 +263,7 @@ CORS is same-origin by default when requests enter through nginx on port 2026. S
 | **Uploads** (`/api/threads/{id}/uploads`) | `POST /` - upload files (auto-converts PDF/PPT/Excel/Word); `GET /list` - list; `DELETE /{filename}` - delete |
 | **Threads** (`/api/threads/{id}`) | `DELETE /` - remove DeerFlow-managed local thread data after LangGraph thread deletion; unexpected failures are logged server-side and return a generic 500 detail |
 | **Artifacts** (`/api/threads/{id}/artifacts`) | `GET /{path}` - serve artifacts; active content types (`text/html`, `application/xhtml+xml`, `image/svg+xml`) are always forced as download attachments to reduce XSS risk; `?download=true` still forces download for other file types |
-| **Suggestions** (`/api/threads/{id}/suggestions`) | `POST /` - generate follow-up questions; rich list/block model content is normalized before JSON parsing |
+| **Suggestions** (`/api/threads/{id}/suggestions`) | `POST /` - generate follow-up questions; rich list/block model content is normalized and inline reasoning (`<think>...</think>`, including unclosed/truncated blocks from reasoning models like MiniMax-M3) is stripped before JSON parsing |
 | **Thread Runs** (`/api/threads/{id}/runs`) | `POST /` - create background run; `POST /stream` - create + SSE stream; `POST /wait` - create + block; `GET /` - list runs; `GET /{rid}` - run details; `POST /{rid}/cancel` - cancel; `GET /{rid}/join` - join SSE; `GET /{rid}/messages` - paginated messages `{data, has_more}`; `GET /{rid}/events` - full event stream; `GET /../messages` - thread messages with feedback; `GET /../token-usage` - aggregate tokens |
 | **Feedback** (`/api/threads/{id}/runs/{rid}/feedback`) | `PUT /` - upsert feedback; `DELETE /` - delete user feedback; `POST /` - create feedback; `GET /` - list feedback; `GET /stats` - aggregate stats; `DELETE /{fid}` - delete specific |
 | **Runs** (`/api/runs`) | `POST /stream` - stateless run + SSE; `POST /wait` - stateless run + block; `GET /{rid}/messages` - paginated messages by run_id `{data, has_more}` (cursor: `after_seq`/`before_seq`); `GET /{rid}/feedback` - list feedback by run_id |
diff --git a/backend/app/gateway/routers/suggestions.py b/backend/app/gateway/routers/suggestions.py
index 56d99b9d3..39f7d250a 100644
--- a/backend/app/gateway/routers/suggestions.py
+++ b/backend/app/gateway/routers/suggestions.py
@@ -1,5 +1,6 @@
 import json
 import logging
+import re
 
 from fastapi import APIRouter, Depends, Request
 from langchain_core.messages import HumanMessage, SystemMessage
@@ -30,6 +31,31 @@ class SuggestionsResponse(BaseModel):
     suggestions: list[str] = Field(default_factory=list, description="Suggested follow-up questions")
 
 
+# Matches a complete <think>...</think> block (case-insensitive, spans newlines).
+_THINK_BLOCK_RE = re.compile(r"<think\b[^>]*>.*?</think\s*>", re.IGNORECASE | re.DOTALL)
+# Matches a dangling, unclosed <think> (model truncated at max_tokens mid-thought).
+_OPEN_THINK_RE = re.compile(r"<think\b[^>]*>", re.IGNORECASE)
+
+
+def _strip_think_blocks(text: str) -> str:
+    """Remove reasoning-model ``<think>...</think>`` blocks from the response.
+
+    Reasoning models such as MiniMax-M3 inline their chain-of-thought into the
+    message ``content`` wrapped in ``<think>...</think>`` (``reasoning_split``
+    defaults to false), rather than exposing a separate ``reasoning_content``
+    field. The thinking text frequently contains ``[`` / ``]`` characters, which
+    corrupted the downstream ``find('[')`` / ``rfind(']')`` JSON extraction and
+    produced empty suggestions. We strip the reasoning before parsing so only
+    the actual answer remains.
+    """
+    text = _THINK_BLOCK_RE.sub("", text)
+    # Drop any unclosed <think> (and everything after it) left by truncation.
+    open_match = _OPEN_THINK_RE.search(text)
+    if open_match:
+        text = text[: open_match.start()]
+    return text.strip()
+
+
 def _strip_markdown_code_fence(text: str) -> str:
     stripped = text.strip()
     if not stripped.startswith("```"):
@@ -41,7 +67,8 @@ def _strip_markdown_code_fence(text: str) -> str:
 
 
 def _parse_json_string_list(text: str) -> list[str] | None:
-    candidate = _strip_markdown_code_fence(text)
+    candidate = _strip_think_blocks(text)
+    candidate = _strip_markdown_code_fence(candidate)
     start = candidate.find("[")
     end = candidate.rfind("]")
     if start == -1 or end == -1 or end <= start:
diff --git a/backend/tests/test_suggestions_router.py b/backend/tests/test_suggestions_router.py
index 0058e4588..bd0a998ff 100644
--- a/backend/tests/test_suggestions_router.py
+++ b/backend/tests/test_suggestions_router.py
@@ -25,6 +25,60 @@ def test_parse_json_string_list_rejects_non_list():
     assert suggestions._parse_json_string_list(text) is None
 
 
+def test_strip_think_blocks_removes_complete_block():
+    text = "<think>\nreasoning here\n</think>\nanswer"
+    assert suggestions._strip_think_blocks(text) == "answer"
+
+
+def test_strip_think_blocks_is_case_insensitive():
+    text = "<Think>reasoning</THINK>\nanswer"
+    assert suggestions._strip_think_blocks(text) == "answer"
+
+
+def test_strip_think_blocks_drops_unclosed_block():
+    # Reasoning models truncated at max_tokens emit an unclosed <think>.
+    text = "<think>\nreasoning that never finished because tokens ran out"
+    assert suggestions._strip_think_blocks(text) == ""
+
+
+def test_strip_think_blocks_keeps_text_without_think():
+    text = '["a", "b"]'
+    assert suggestions._strip_think_blocks(text) == '["a", "b"]'
+
+
+def test_parse_json_string_list_ignores_brackets_inside_think_block():
+    # MiniMax-M3 inlines its chain-of-thought as <think>...</think> in content
+    # (reasoning_split=false). When that reasoning contains '[' / ']', the old
+    # find('[')/rfind(']') logic grabbed the wrong span and parsing failed.
+    text = '<think>\nMaybe a list like ["x", "y"] could work. Let me craft 3.\n</think>\n["Q1", "Q2", "Q3"]'
+    assert suggestions._parse_json_string_list(text) == ["Q1", "Q2", "Q3"]
+
+
+def test_parse_json_string_list_strips_think_then_code_fence():
+    text = '<think>reasoning</think>\n```json\n["Q1", "Q2"]\n```'
+    assert suggestions._parse_json_string_list(text) == ["Q1", "Q2"]
+
+
+def test_generate_suggestions_strips_inline_think_block(monkeypatch):
+    # End-to-end: model returns thinking inline followed by the JSON array.
+    req = suggestions.SuggestionsRequest(
+        messages=[
+            suggestions.SuggestionMessage(role="user", content="介绍深度学习"),
+            suggestions.SuggestionMessage(role="assistant", content="深度学习是机器学习的分支。"),
+        ],
+        n=3,
+        model_name=None,
+    )
+    content = '<think>\nThe user asked about deep learning. Options: maybe [1] frameworks, [2] math basics.\n</think>\n["深度学习和机器学习的区别？", "常用框架有哪些？", "需要什么数学基础？"]'
+    fake_model = MagicMock()
+    fake_model.ainvoke = AsyncMock(return_value=MagicMock(content=content))
+    monkeypatch.setattr(suggestions, "create_chat_model", lambda **kwargs: fake_model)
+
+    result = asyncio.run(suggestions.generate_suggestions.__wrapped__("t1", req, request=None, config=SimpleNamespace()))
+
+    assert result.suggestions == ["深度学习和机器学习的区别？", "常用框架有哪些？", "需要什么数学基础？"]
+
+
 def test_format_conversation_formats_roles():
     messages = [
         suggestions.SuggestionMessage(role="User", content="Hi"),
diff --git a/config.example.yaml b/config.example.yaml
index 0a0026d87..99752cf5e 100644
--- a/config.example.yaml
+++ b/config.example.yaml
@@ -289,7 +289,23 @@ models:
   #   temperature: 1.0  # MiniMax requires temperature in (0.0, 1.0]
   #   supports_vision: true
   #   supports_thinking: true
+  #   # MiniMax inlines its chain-of-thought into `content` as <think>...</think>
+  #   # (reasoning_split defaults to false), not in a separate reasoning_content
+  #   # field. Declare the thinking toggle so non-thinking paths (flash mode,
+  #   # follow-up suggestions, title/memory generation) truly disable reasoning
+  #   # instead of wasting tokens on — and parsing around — inline <think> blocks.
+  #   when_thinking_enabled:
+  #     extra_body:
+  #       thinking:
+  #         type: adaptive
+  #   when_thinking_disabled:
+  #     extra_body:
+  #       thinking:
+  #         type: disabled
 
+  # NOTE: M2.x models always think — passing thinking:{type:disabled} has no
+  # effect (per MiniMax docs), so the toggle above is omitted for M2.7. The
+  # follow-up-suggestions endpoint strips inline <think> defensively regardless.
   # - name: minimax-m2.7
   #   display_name: MiniMax M2.7
   #   use: langchain_openai:ChatOpenAI
@@ -331,7 +347,23 @@ models:
   #   temperature: 1.0  # MiniMax requires temperature in (0.0, 1.0]
   #   supports_vision: true
   #   supports_thinking: true
+  #   # MiniMax inlines its chain-of-thought into `content` as <think>...</think>
+  #   # (reasoning_split defaults to false), not in a separate reasoning_content
+  #   # field. Declare the thinking toggle so non-thinking paths (flash mode,
+  #   # follow-up suggestions, title/memory generation) truly disable reasoning
+  #   # instead of wasting tokens on — and parsing around — inline <think> blocks.
+  #   when_thinking_enabled:
+  #     extra_body:
+  #       thinking:
+  #         type: adaptive
+  #   when_thinking_disabled:
+  #     extra_body:
+  #       thinking:
+  #         type: disabled
 
+  # NOTE: M2.x models always think — passing thinking:{type:disabled} has no
+  # effect (per MiniMax docs), so the toggle above is omitted for M2.7. The
+  # follow-up-suggestions endpoint strips inline <think> defensively regardless.
   # - name: minimax-m2.7
   #   display_name: MiniMax M2.7
   #   use: langchain_openai:ChatOpenAI