diff --git a/README.md b/README.md
index 5a2c1dc..534e942 100644
--- a/README.md
+++ b/README.md
@@ -324,6 +324,7 @@ Optional CLI flags (API):
 | **OpenRouter** | `openrouter.ai/api/v1` | Supports Gemini/Claude/others |
 | **Bianxie** | `api.bianxie.ai/v1` | OpenAI-compatible API |
 | **Gemini (Google)** | `generativelanguage.googleapis.com/v1beta` | Official Google Gemini API (`google-genai`) |
+| **MiniMax** | `api.minimax.io/v1` | OpenAI-compatible API; SVG generation only (no image gen) |
 
 Common CLI flags:
 
@@ -337,6 +338,33 @@ Common CLI flags:
 - `--merge_threshold` (0 disables merging)
 - `--optimize_iterations` (0 disables optimization)
 - `--reference_image_path` (optional)
+- `--figure_path` (pre-generated figure image, skips step 1)
+
+### Using MiniMax
+
+[MiniMax](https://www.minimaxi.com/) provides OpenAI-compatible LLM APIs with strong multimodal capabilities (MiniMax-M2.7, 204K context). MiniMax excels at SVG generation and text understanding tasks but does not support image generation. Use `--figure_path` to provide a pre-generated figure:
+
+```bash
+# Step 1: Generate figure with another provider
+python autofigure2.py \
+  --method_file paper.txt \
+  --output_dir outputs/step1 \
+  --provider gemini \
+  --api_key YOUR_GEMINI_KEY \
+  --stop_after 1
+
+# Step 2: Use MiniMax for SVG generation (steps 2-5)
+export MINIMAX_API_KEY="your-minimax-key"
+python autofigure2.py \
+  --method_file paper.txt \
+  --output_dir outputs/demo \
+  --provider minimax \
+  --figure_path outputs/step1/figure.png
+```
+
+Available MiniMax models:
+- `MiniMax-M2.7` (default) — latest model, 204K context
+- `MiniMax-M2.7-highspeed` — faster variant, 204K context
 
 ### Custom Provider / Custom Base URL
 
diff --git a/README_ZH.md b/README_ZH.md
index baf13ef..b35374f 100644
--- a/README_ZH.md
+++ b/README_ZH.md
@@ -314,10 +314,11 @@ python autofigure2.py \
 | **OpenRouter** | `openrouter.ai/api/v1` | 支持 Gemini/Claude/其他模型 |
 | **Bianxie** | `api.bianxie.ai/v1` | 兼容 OpenAI 接口 |
 | **Gemini (Google)** | `generativelanguage.googleapis.com/v1beta` | Google 官方 Gemini API（`google-genai`） |
+| **MiniMax** | `api.minimax.io/v1` | 兼容 OpenAI 接口；仅 SVG 生成（不支持图像生成） |
 
 常用 CLI 参数：
 
-- `--provider` (openrouter | bianxie | gemini)
+- `--provider` (openrouter | bianxie | gemini | minimax)
 - `--image_model`, `--svg_model`
 - `--image_size` (1K | 2K | 4K，仅 Gemini)
 - `--sam_prompt` (逗号分隔的提示词)
@@ -327,6 +328,33 @@ python autofigure2.py \
 - `--merge_threshold` (0 禁用合并)
 - `--optimize_iterations` (0 禁用优化)
 - `--reference_image_path` (可选)
+- `--figure_path` (预生成图片路径，跳过步骤一)
+
+### 使用 MiniMax
+
+[MiniMax](https://www.minimaxi.com/) 提供兼容 OpenAI 的 LLM API，具有强大的多模态能力（MiniMax-M2.7，204K 上下文）。MiniMax 擅长 SVG 生成和文本理解，但不支持图像生成。使用 `--figure_path` 提供预生成的图片：
+
+```bash
+# 步骤 1：用其他 provider 生成图片
+python autofigure2.py \
+  --method_file paper.txt \
+  --output_dir outputs/step1 \
+  --provider gemini \
+  --api_key YOUR_GEMINI_KEY \
+  --stop_after 1
+
+# 步骤 2：用 MiniMax 进行 SVG 生成（步骤 2-5）
+export MINIMAX_API_KEY="your-minimax-key"
+python autofigure2.py \
+  --method_file paper.txt \
+  --output_dir outputs/demo \
+  --provider minimax \
+  --figure_path outputs/step1/figure.png
+```
+
+可用的 MiniMax 模型：
+- `MiniMax-M2.7`（默认）— 最新模型，204K 上下文
+- `MiniMax-M2.7-highspeed` — 高速版本，204K 上下文
 
 ### 自定义提供商 / 自定义 Base URL
 
diff --git a/autofigure2.py b/autofigure2.py
index 8101b7f..ec5188a 100644
--- a/autofigure2.py
+++ b/autofigure2.py
@@ -5,6 +5,7 @@
 - openrouter: OpenRouter API (https://openrouter.ai/api/v1)
 - bianxie: Bianxie API (https://api.bianxie.ai/v1) - 使用 OpenAI SDK
 - gemini: Google Gemini 官方 API (https://ai.google.dev/)
+- minimax: MiniMax API (https://api.minimax.io/v1) - 使用 OpenAI SDK（仅 SVG 生成，不支持图像生成）
 
 占位符模式 (--placeholder_mode):
 - none: 无特殊样式（默认黑色边框）
@@ -104,9 +105,14 @@
         "default_image_model": "gemini-3-pro-image-preview",
         "default_svg_model": "gemini-3.1-pro",
     },
+    "minimax": {
+        "base_url": "https://api.minimax.io/v1",
+        "default_image_model": None,
+        "default_svg_model": "MiniMax-M2.7",
+    },
 }
 
-ProviderType = Literal["openrouter", "bianxie", "gemini"]
+ProviderType = Literal["openrouter", "bianxie", "gemini", "minimax"]
 PlaceholderMode = Literal["none", "box", "label"]
 GEMINI_DEFAULT_IMAGE_SIZE = "4K"
 IMAGE_SIZE_CHOICES = ("1K", "2K", "4K")
@@ -158,6 +164,8 @@ def call_llm_text(
         return _call_bianxie_text(prompt, api_key, model, base_url, max_tokens, temperature)
     if provider == "gemini":
         return _call_gemini_text(prompt, api_key, model, max_tokens, temperature)
+    if provider == "minimax":
+        return _call_minimax_text(prompt, api_key, model, base_url, max_tokens, temperature)
     return _call_openrouter_text(prompt, api_key, model, base_url, max_tokens, temperature)
 
 
@@ -189,6 +197,8 @@ def call_llm_multimodal(
         return _call_bianxie_multimodal(contents, api_key, model, base_url, max_tokens, temperature)
     if provider == "gemini":
         return _call_gemini_multimodal(contents, api_key, model, max_tokens, temperature)
+    if provider == "minimax":
+        return _call_minimax_multimodal(contents, api_key, model, base_url, max_tokens, temperature)
     return _call_openrouter_multimodal(contents, api_key, model, base_url, max_tokens, temperature)
 
 
@@ -224,6 +234,8 @@ def call_llm_image_generation(
             reference_image=reference_image,
             image_size=image_size,
         )
+    if provider == "minimax":
+        return _call_minimax_image_generation(prompt, api_key, model, base_url, reference_image)
     return _call_openrouter_image_generation(prompt, api_key, model, base_url, reference_image)
 
 
@@ -941,6 +953,114 @@ def _call_gemini_image_generation(
         raise
 
 
+# ============================================================================
+# MiniMax Provider 实现 (OpenAI-compatible API)
+# ============================================================================
+
+def _clamp_minimax_temperature(temperature: float) -> float:
+    """MiniMax requires temperature in (0.0, 1.0]."""
+    if temperature <= 0.0:
+        return 0.01
+    if temperature > 1.0:
+        return 1.0
+    return temperature
+
+
+def _call_minimax_text(
+    prompt: str,
+    api_key: str,
+    model: str,
+    base_url: str,
+    max_tokens: int = 16000,
+    temperature: float = 0.7,
+) -> Optional[str]:
+    """Call MiniMax text API via OpenAI SDK (OpenAI-compatible endpoint)."""
+    try:
+        from openai import OpenAI
+
+        client = OpenAI(base_url=base_url, api_key=api_key)
+        temperature = _clamp_minimax_temperature(temperature)
+
+        completion = client.chat.completions.create(
+            model=model,
+            messages=[{"role": "user", "content": prompt}],
+            max_tokens=max_tokens,
+            temperature=temperature,
+        )
+
+        return completion.choices[0].message.content if completion and completion.choices else None
+    except Exception as e:
+        print(f"[MiniMax] API call failed: {e}")
+        raise
+
+
+def _call_minimax_multimodal(
+    contents: List[Any],
+    api_key: str,
+    model: str,
+    base_url: str,
+    max_tokens: int = 16000,
+    temperature: float = 0.7,
+) -> Optional[str]:
+    """Call MiniMax multimodal API via OpenAI SDK (vision support)."""
+    try:
+        from openai import OpenAI
+
+        client = OpenAI(base_url=base_url, api_key=api_key)
+        temperature = _clamp_minimax_temperature(temperature)
+
+        message_content: List[Dict[str, Any]] = []
+        for part in contents:
+            if isinstance(part, str):
+                message_content.append({"type": "text", "text": part})
+            elif isinstance(part, Image.Image):
+                buf = io.BytesIO()
+                part.save(buf, format='PNG')
+                image_b64 = base64.b64encode(buf.getvalue()).decode('utf-8')
+                message_content.append({
+                    "type": "image_url",
+                    "image_url": {"url": f"data:image/png;base64,{image_b64}"}
+                })
+
+        completion = client.chat.completions.create(
+            model=model,
+            messages=[{"role": "user", "content": message_content}],
+            max_tokens=max_tokens,
+            temperature=temperature,
+        )
+
+        text = completion.choices[0].message.content if completion and completion.choices else None
+        if text:
+            # Strip <think>...</think> tags from MiniMax reasoning models,
+            # but only if there is content outside the think block.
+            stripped = re.sub(r"<think>.*?</think>", "", text, flags=re.DOTALL).strip()
+            if stripped:
+                text = stripped
+        return text
+    except Exception as e:
+        print(f"[MiniMax] Multimodal API call failed: {e}")
+        raise
+
+
+def _call_minimax_image_generation(
+    prompt: str,
+    api_key: str,
+    model: str,
+    base_url: str,
+    reference_image: Optional[Image.Image] = None,
+) -> Optional[Image.Image]:
+    """MiniMax does not support image generation.
+
+    Use --figure_path to provide a pre-generated figure when using
+    MiniMax as the provider, or use a different provider for step 1.
+    """
+    raise NotImplementedError(
+        "MiniMax does not support image generation. "
+        "Use --figure_path to provide a pre-generated figure, "
+        "or use a different provider (openrouter/bianxie/gemini) for image generation."
+    )
+
+
 # ============================================================================
 # 步骤一：调用 LLM 生成图片
 # ============================================================================
@@ -2854,6 +2974,7 @@ def method_to_svg(
     optimize_iterations: int = 2,
     merge_threshold: float = 0.9,
     image_size: str = GEMINI_DEFAULT_IMAGE_SIZE,
+    figure_path: Optional[str] = None,
 ) -> dict:
     """
     完整流程：Paper Method → SVG with Icons
@@ -2879,6 +3000,7 @@ def method_to_svg(
             - "label": 灰色填充+黑色边框+序号标签（推荐）
         optimize_iterations: 步骤 4.6 优化迭代次数（0 表示跳过优化）
         merge_threshold: Box合并阈值，重叠比例超过此值则合并（0表示不合并，默认0.9）
+        figure_path: 预生成的图片路径（跳过步骤一），适用于不支持图像生成的 provider（如 MiniMax）
 
     Returns:
         结果字典
@@ -2919,24 +3041,35 @@ def method_to_svg(
         print(f"生图分辨率: {image_size}")
     print("=" * 60)
 
-    # 步骤一：生成图片
-    figure_path = output_dir / "figure.png"
-    generate_figure_from_method(
-        method_text=method_text,
-        output_path=str(figure_path),
-        api_key=api_key,
-        model=image_gen_model,
-        base_url=base_url,
-        provider=provider,
-        image_size=image_size,
-    )
+    # 步骤一：生成图片（或使用预生成的图片）
+    figure_path_out = output_dir / "figure.png"
+    if figure_path:
+        # 使用预生成的图片，跳过 LLM 图像生成
+        src = Path(figure_path)
+        if not src.is_file():
+            raise FileNotFoundError(f"预生成的图片不存在: {figure_path}")
+        print("=" * 60)
+        print("步骤一：使用预生成图片（跳过 LLM 图像生成）")
+        print("=" * 60)
+        print(f"图片来源: {figure_path}")
+        shutil.copy2(str(src), str(figure_path_out))
+    else:
+        generate_figure_from_method(
+            method_text=method_text,
+            output_path=str(figure_path_out),
+            api_key=api_key,
+            model=image_gen_model,
+            base_url=base_url,
+            provider=provider,
+            image_size=image_size,
+        )
 
     if stop_after == 1:
         print("\n" + "=" * 60)
         print("已在步骤 1 后停止")
         print("=" * 60)
         return {
-            "figure_path": str(figure_path),
+            "figure_path": str(figure_path_out),
             "samed_path": None,
             "boxlib_path": None,
             "icon_infos": [],
@@ -2947,7 +3080,7 @@ def method_to_svg(
 
     # 步骤二：SAM3 分割（包含Box合并）
     samed_path, boxlib_path, valid_boxes = segment_with_sam3(
-        image_path=str(figure_path),
+        image_path=str(figure_path_out),
         output_dir=str(output_dir),
         text_prompts=sam_prompts,
         min_score=min_score,
@@ -2968,7 +3101,7 @@ def method_to_svg(
         print("已在步骤 2 后停止")
         print("=" * 60)
         return {
-            "figure_path": str(figure_path),
+            "figure_path": str(figure_path_out),
             "samed_path": samed_path,
             "boxlib_path": boxlib_path,
             "icon_infos": [],
@@ -2984,7 +3117,7 @@ def method_to_svg(
     else:
         _ensure_rmbg2_access_ready(rmbg_model_path)
         icon_infos = crop_and_remove_background(
-            image_path=str(figure_path),
+            image_path=str(figure_path_out),
             boxlib_path=boxlib_path,
             output_dir=str(output_dir),
             rmbg_model_path=rmbg_model_path,
@@ -2995,7 +3128,7 @@ def method_to_svg(
         print("已在步骤 3 后停止")
         print("=" * 60)
         return {
-            "figure_path": str(figure_path),
+            "figure_path": str(figure_path_out),
             "samed_path": samed_path,
             "boxlib_path": boxlib_path,
             "icon_infos": icon_infos,
@@ -3010,7 +3143,7 @@ def method_to_svg(
     final_svg_path = output_dir / "final.svg"
     try:
         generate_svg_template(
-            figure_path=str(figure_path),
+            figure_path=str(figure_path_out),
             samed_path=samed_path,
             boxlib_path=boxlib_path,
             output_path=str(template_svg_path),
@@ -3024,7 +3157,7 @@ def method_to_svg(
 
         # 步骤 4.6：LLM 优化 SVG 模板（可配置迭代次数，0 表示跳过）
         optimize_svg_with_llm(
-            figure_path=str(figure_path),
+            figure_path=str(figure_path_out),
             samed_path=samed_path,
             final_svg_path=str(template_svg_path),
             output_path=str(optimized_template_path),
@@ -3041,7 +3174,7 @@ def method_to_svg(
             raise
         print(f"无图标模式下 SVG 重建失败（{exc}），改用内嵌原图的保底 SVG")
         create_embedded_figure_svg(
-            figure_path=str(figure_path),
+            figure_path=str(figure_path_out),
             output_path=str(final_svg_path),
         )
 
@@ -3050,7 +3183,7 @@ def method_to_svg(
         print("已在步骤 4 后停止")
         print("=" * 60)
         return {
-            "figure_path": str(figure_path),
+            "figure_path": str(figure_path_out),
             "samed_path": samed_path,
             "boxlib_path": boxlib_path,
             "icon_infos": icon_infos,
@@ -3069,7 +3202,7 @@ def method_to_svg(
         else:
             print("无图标模式缺少模板 SVG，生成保底 final.svg")
             create_embedded_figure_svg(
-                figure_path=str(figure_path),
+                figure_path=str(figure_path_out),
                 output_path=str(final_svg_path),
             )
     else:
@@ -3078,7 +3211,7 @@ def method_to_svg(
         print("步骤 4.7：坐标系对齐")
         print("-" * 50)
 
-        figure_img = Image.open(figure_path)
+        figure_img = Image.open(figure_path_out)
         figure_width, figure_height = figure_img.size
         print(f"原图尺寸: {figure_width} x {figure_height}")
 
@@ -3114,7 +3247,7 @@ def method_to_svg(
     print("\n" + "=" * 60)
     print("流程完成！")
     print("=" * 60)
-    print(f"原始图片: {figure_path}")
+    print(f"原始图片: {figure_path_out}")
     print(f"标记图片: {samed_path}")
     print(f"Box信息: {boxlib_path}")
     print(f"图标数量: {len(icon_infos)}")
@@ -3123,7 +3256,7 @@ def method_to_svg(
     print(f"最终SVG: {final_svg_path}")
 
     return {
-        "figure_path": str(figure_path),
+        "figure_path": str(figure_path_out),
         "samed_path": samed_path,
         "boxlib_path": boxlib_path,
         "icon_infos": icon_infos,
@@ -3180,13 +3313,13 @@ def create_embedded_figure_svg(
     # Provider 参数
     parser.add_argument(
         "--provider",
-        choices=["openrouter", "bianxie", "gemini"],
+        choices=["openrouter", "bianxie", "gemini", "minimax"],
         default="bianxie",
         help="API 提供商（默认: bianxie）"
     )
 
     # API 参数
-    parser.add_argument("--api_key", default=None, help="API Key")
+    parser.add_argument("--api_key", default=None, help="API Key（或设置 MINIMAX_API_KEY 环境变量）")
     parser.add_argument("--base_url", default=None, help="API base URL（默认根据 provider 自动设置）")
 
     # 模型参数
@@ -3199,6 +3332,13 @@ def create_embedded_figure_svg(
     )
     parser.add_argument("--svg_model", default=None, help="SVG生成模型（默认根据 provider 自动设置）")
 
+    # 预生成图片参数（跳过步骤一，适用于 MiniMax 等不支持图像生成的 provider）
+    parser.add_argument(
+        "--figure_path",
+        default=None,
+        help="预生成的图片路径（跳过步骤一的 LLM 图像生成，适用于 minimax 等不支持图像生成的 provider）",
+    )
+
     # Step 1 参考图片参数
     parser.add_argument(
         "--use_reference_image",
@@ -3266,12 +3406,19 @@ def create_embedded_figure_svg(
         parser.error("--use_reference_image 需要 --reference_image_path")
     if args.reference_image_path and not Path(args.reference_image_path).is_file():
         parser.error(f"参考图片不存在: {args.reference_image_path}")
+    if args.figure_path and not Path(args.figure_path).is_file():
+        parser.error(f"预生成图片不存在: {args.figure_path}")
 
     USE_REFERENCE_IMAGE = bool(args.use_reference_image)
     REFERENCE_IMAGE_PATH = args.reference_image_path
     if REFERENCE_IMAGE_PATH:
         USE_REFERENCE_IMAGE = True
 
+    # API key: CLI flag > environment variable
+    api_key = args.api_key
+    if not api_key and args.provider == "minimax":
+        api_key = os.environ.get("MINIMAX_API_KEY")
+
     # 获取 method 文本：优先使用 --method_text
     method_text = args.method_text
     if method_text is None:
@@ -3282,7 +3429,7 @@ def create_embedded_figure_svg(
     result = method_to_svg(
         method_text=method_text,
         output_dir=args.output_dir,
-        api_key=args.api_key,
+        api_key=api_key,
         base_url=args.base_url,
         provider=args.provider,
         image_gen_model=args.image_model,
@@ -3298,4 +3445,5 @@ def create_embedded_figure_svg(
         placeholder_mode=args.placeholder_mode,
         optimize_iterations=args.optimize_iterations,
         merge_threshold=args.merge_threshold,
+        figure_path=args.figure_path,
     )
diff --git a/server.py b/server.py
index 4cecf5c..13a5502 100644
--- a/server.py
+++ b/server.py
@@ -101,6 +101,7 @@ class RunRequest(BaseModel):
     merge_threshold: Optional[float] = None
     optimize_iterations: Optional[int] = None
     reference_image_path: Optional[str] = None
+    figure_path: Optional[str] = None
 
 
 app = FastAPI()
@@ -174,6 +175,15 @@ def run_job(req: RunRequest) -> JSONResponse:
         )
         cmd += ["--reference_image_path", reference_path]
 
+    if req.figure_path:
+        fig_path = req.figure_path
+        fig_path = (
+            str((BASE_DIR / fig_path).resolve())
+            if not Path(fig_path).is_absolute()
+            else fig_path
+        )
+        cmd += ["--figure_path", fig_path]
+
     env = os.environ.copy()
     env["PYTHONUNBUFFERED"] = "1"
 
diff --git a/tests/__init__.py b/tests/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/tests/test_minimax_provider.py b/tests/test_minimax_provider.py
new file mode 100644
index 0000000..f6077a4
--- /dev/null
+++ b/tests/test_minimax_provider.py
@@ -0,0 +1,357 @@
+"""Unit tests for the MiniMax LLM provider integration."""
+
+from __future__ import annotations
+
+import io
+import json
+import os
+import shutil
+import tempfile
+from pathlib import Path
+from typing import Any
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+# ---------------------------------------------------------------------------
+# Import helpers – the module under test uses heavy ML deps (torch, etc.)
+# We mock them when they are not installed so the provider-level tests still
+# run in lightweight CI environments.
+# ---------------------------------------------------------------------------
+
+import importlib
+import sys
+
+_STUBS: dict[str, Any] = {}
+
+
+def _ensure_stubs():
+    """Create lightweight stubs for heavy optional deps if they are missing."""
+    for mod_name in ("torch", "torchvision", "torchvision.transforms",
+                     "timm", "transformers", "kornia"):
+        if mod_name not in sys.modules:
+            stub = MagicMock()
+            sys.modules[mod_name] = stub
+            _STUBS[mod_name] = stub
+
+
+_ensure_stubs()
+
+# Now we can import the target module
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+import autofigure2 as af
+
+
+# ============================================================================
+# Temperature clamping
+# ============================================================================
+
+class TestMiniMaxTemperatureClamping:
+    """_clamp_minimax_temperature must keep values in (0, 1]."""
+
+    def test_zero_is_clamped_up(self):
+        assert af._clamp_minimax_temperature(0.0) == 0.01
+
+    def test_negative_is_clamped_up(self):
+        assert af._clamp_minimax_temperature(-0.5) == 0.01
+
+    def test_above_one_is_clamped_down(self):
+        assert af._clamp_minimax_temperature(1.5) == 1.0
+
+    def test_exactly_one_is_kept(self):
+        assert af._clamp_minimax_temperature(1.0) == 1.0
+
+    def test_normal_value_is_kept(self):
+        assert af._clamp_minimax_temperature(0.7) == 0.7
+
+    def test_small_positive_is_kept(self):
+        assert af._clamp_minimax_temperature(0.01) == 0.01
+
+
+# ============================================================================
+# Provider configuration
+# ============================================================================
+
+class TestMiniMaxProviderConfig:
+    """PROVIDER_CONFIGS must include minimax with correct defaults."""
+
+    def test_minimax_in_configs(self):
+        assert "minimax" in af.PROVIDER_CONFIGS
+
+    def test_base_url(self):
+        assert af.PROVIDER_CONFIGS["minimax"]["base_url"] == "https://api.minimax.io/v1"
+
+    def test_default_svg_model(self):
+        assert af.PROVIDER_CONFIGS["minimax"]["default_svg_model"] == "MiniMax-M2.7"
+
+    def test_no_default_image_model(self):
+        assert af.PROVIDER_CONFIGS["minimax"]["default_image_model"] is None
+
+
+# ============================================================================
+# Image generation raises NotImplementedError
+# ============================================================================
+
+class TestMiniMaxImageGeneration:
+    """MiniMax must refuse image generation with a clear error."""
+
+    def test_raises_not_implemented(self):
+        with pytest.raises(NotImplementedError, match="MiniMax does not support image generation"):
+            af._call_minimax_image_generation(
+                prompt="draw a cat",
+                api_key="test-key",
+                model="MiniMax-M2.7",
+                base_url="https://api.minimax.io/v1",
+            )
+
+    def test_dispatcher_raises_for_minimax(self):
+        with pytest.raises(NotImplementedError):
+            af.call_llm_image_generation(
+                prompt="draw a cat",
+                api_key="test-key",
+                model="MiniMax-M2.7",
+                base_url="https://api.minimax.io/v1",
+                provider="minimax",
+            )
+
+
+# ============================================================================
+# Text call dispatching
+# ============================================================================
+
+class TestMiniMaxTextCall:
+    """call_llm_text must route to _call_minimax_text for minimax provider."""
+
+    @patch("autofigure2._call_minimax_text", return_value="hello from minimax")
+    def test_dispatch(self, mock_call):
+        result = af.call_llm_text(
+            prompt="say hello",
+            api_key="test-key",
+            model="MiniMax-M2.7",
+            base_url="https://api.minimax.io/v1",
+            provider="minimax",
+        )
+        assert result == "hello from minimax"
+        mock_call.assert_called_once()
+
+    @patch("autofigure2._call_minimax_text")
+    def test_temperature_passed(self, mock_call):
+        af.call_llm_text(
+            prompt="hi",
+            api_key="k",
+            model="MiniMax-M2.7",
+            base_url="https://api.minimax.io/v1",
+            provider="minimax",
+            temperature=0.5,
+        )
+        _, kwargs = mock_call.call_args
+        assert kwargs.get("temperature", mock_call.call_args[0][5] if len(mock_call.call_args[0]) > 5 else None) is not None
+
+
+# ============================================================================
+# Multimodal call dispatching
+# ============================================================================
+
+class TestMiniMaxMultimodalCall:
+    """call_llm_multimodal must route to _call_minimax_multimodal."""
+
+    @patch("autofigure2._call_minimax_multimodal", return_value="svg code here")
+    def test_dispatch(self, mock_call):
+        result = af.call_llm_multimodal(
+            contents=["generate svg"],
+            api_key="test-key",
+            model="MiniMax-M2.7",
+            base_url="https://api.minimax.io/v1",
+            provider="minimax",
+        )
+        assert result == "svg code here"
+        mock_call.assert_called_once()
+
+
+# ============================================================================
+# Think-tag stripping in multimodal response
+# ============================================================================
+
+class TestThinkTagStripping:
+    """MiniMax multimodal must strip <think> tags from responses."""
+
+    @patch("autofigure2.OpenAI", create=True)
+    def test_think_tag_stripped_when_content_outside(self, mock_openai_cls):
+        """Think tags should be stripped when there is content outside them."""
+        mock_msg = MagicMock()
+        mock_msg.content = "<think>reasoning here</think>SVG output"
+        mock_choice = MagicMock()
+        mock_choice.message = mock_msg
+        mock_completion = MagicMock()
+        mock_completion.choices = [mock_choice]
+
+        mock_client = MagicMock()
+        mock_client.chat.completions.create.return_value = mock_completion
+        mock_openai_cls.return_value = mock_client
+
+        with patch.dict("sys.modules", {"openai": MagicMock(OpenAI=mock_openai_cls)}):
+            result = af._call_minimax_multimodal(
+                contents=["test"],
+                api_key="key",
+                model="MiniMax-M2.7",
+                base_url="https://api.minimax.io/v1",
+            )
+
+        assert "<think>" not in result
+        assert result == "SVG output"
+
+    @patch("autofigure2.OpenAI", create=True)
+    def test_think_tag_preserved_when_only_content(self, mock_openai_cls):
+        """If all content is inside think tags, preserve the original response."""
+        mock_msg = MagicMock()
+        mock_msg.content = "<think>the entire answer is here</think>"
+        mock_choice = MagicMock()
+        mock_choice.message = mock_msg
+        mock_completion = MagicMock()
+        mock_completion.choices = [mock_choice]
+
+        mock_client = MagicMock()
+        mock_client.chat.completions.create.return_value = mock_completion
+        mock_openai_cls.return_value = mock_client
+
+        with patch.dict("sys.modules", {"openai": MagicMock(OpenAI=mock_openai_cls)}):
+            result = af._call_minimax_multimodal(
+                contents=["test"],
+                api_key="key",
+                model="MiniMax-M2.7",
+                base_url="https://api.minimax.io/v1",
+            )
+
+        # Should preserve the original since stripping would leave empty
+        assert result == "<think>the entire answer is here</think>"
+
+
+# ============================================================================
+# figure_path support in method_to_svg
+# ============================================================================
+
+class TestFigurePathSupport:
+    """method_to_svg should accept figure_path and skip image generation."""
+
+    def test_figure_path_copies_to_output(self):
+        """When figure_path is provided, step 1 should copy it instead of calling LLM."""
+        # Create a temporary figure file
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Create a minimal PNG (1x1 pixel)
+            from PIL import Image
+            fig_src = Path(tmpdir) / "source_figure.png"
+            img = Image.new("RGB", (100, 100), color="red")
+            img.save(str(fig_src))
+
+            output_dir = Path(tmpdir) / "output"
+            output_dir.mkdir()
+
+            # Patch heavy functions that would run in step 2+
+            with patch.object(af, "segment_with_sam3") as mock_sam, \
+                 patch.object(af, "_ensure_rmbg2_access_ready"), \
+                 patch.object(af, "crop_and_remove_background", return_value=[]), \
+                 patch.object(af, "generate_svg_template"), \
+                 patch.object(af, "optimize_svg_with_llm"), \
+                 patch.object(af, "replace_icons_in_svg"):
+
+                mock_sam.return_value = (str(output_dir / "samed.png"), str(output_dir / "boxlib.json"), [])
+
+                # Create dummy samed.png and boxlib.json that steps expect
+                Image.new("RGB", (100, 100)).save(str(output_dir / "samed.png"))
+                Path(output_dir / "boxlib.json").write_text("[]")
+
+                result = af.method_to_svg(
+                    method_text="test method",
+                    output_dir=str(output_dir),
+                    api_key="test-key",
+                    provider="minimax",
+                    figure_path=str(fig_src),
+                    stop_after=1,
+                )
+
+            # The figure should be copied to output
+            assert (output_dir / "figure.png").is_file()
+            assert result["figure_path"] == str(output_dir / "figure.png")
+
+    def test_missing_figure_path_raises(self):
+        """When figure_path points to a missing file, should raise FileNotFoundError."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            with pytest.raises(FileNotFoundError, match="预生成的图片不存在"):
+                af.method_to_svg(
+                    method_text="test",
+                    output_dir=str(Path(tmpdir) / "output"),
+                    api_key="test-key",
+                    provider="minimax",
+                    figure_path="/nonexistent/figure.png",
+                )
+
+
+# ============================================================================
+# CLI argument parsing
+# ============================================================================
+
+class TestCLIArgs:
+    """Verify minimax is accepted by the CLI argparse config."""
+
+    def test_minimax_in_provider_choices(self):
+        """The --provider choices must include minimax."""
+        import argparse
+        # Extract the provider argument from the parser setup
+        # We check PROVIDER_CONFIGS since the choices list comes from there
+        assert "minimax" in af.PROVIDER_CONFIGS
+
+    def test_figure_path_default_none(self):
+        """--figure_path should default to None when not provided."""
+        # This is implicitly tested by the argparse default
+        # We verify via the method_to_svg signature
+        import inspect
+        sig = inspect.signature(af.method_to_svg)
+        assert sig.parameters["figure_path"].default is None
+
+
+# ============================================================================
+# Integration test (requires MINIMAX_API_KEY)
+# ============================================================================
+
+@pytest.mark.skipif(
+    not os.environ.get("MINIMAX_API_KEY"),
+    reason="MINIMAX_API_KEY not set",
+)
+class TestMiniMaxIntegration:
+    """Integration tests that call the real MiniMax API."""
+
+    def test_text_call(self):
+        """A simple text completion should return a non-empty string."""
+        result = af._call_minimax_text(
+            prompt="Reply with exactly: MINIMAX_OK",
+            api_key=os.environ["MINIMAX_API_KEY"],
+            model="MiniMax-M2.7",
+            base_url="https://api.minimax.io/v1",
+            max_tokens=50,
+            temperature=0.01,
+        )
+        assert result is not None
+        assert len(result) > 0
+
+    def test_multimodal_call(self):
+        """A multimodal call with text-only content should work."""
+        result = af._call_minimax_multimodal(
+            contents=["What is 2+2? Answer with only the number."],
+            api_key=os.environ["MINIMAX_API_KEY"],
+            model="MiniMax-M2.7",
+            base_url="https://api.minimax.io/v1",
+            max_tokens=50,
+            temperature=0.5,
+        )
+        assert result is not None
+        assert len(result) > 0
+
+    def test_image_generation_raises(self):
+        """Image generation must raise NotImplementedError even with a real key."""
+        with pytest.raises(NotImplementedError):
+            af._call_minimax_image_generation(
+                prompt="a red circle",
+                api_key=os.environ["MINIMAX_API_KEY"],
+                model="MiniMax-M2.7",
+                base_url="https://api.minimax.io/v1",
+            )
diff --git a/web/index.html b/web/index.html
index cf479cb..919d21e 100644
--- a/web/index.html
+++ b/web/index.html
@@ -34,6 +34,7 @@
               <option value="bianxie">Bianxie</option>
               <option value="openrouter">OpenRouter</option>
               <option value="gemini" selected>Gemini (Google)</option>
+              <option value="minimax">MiniMax</option>
             </select>
           </div>