diff --git a/Dockerfile b/Dockerfile
index 4792cbd..fa70850 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,4 +1,5 @@
-FROM rocm/vllm:rocm7.12.0_gfx120X-all_ubuntu24.04_py3.12_pytorch_2.9.1_vllm_0.16.0
+# FROM docker.1ms.run/vllm/vllm-openai-rocm:latest
+FROM vllm/vllm-openai-rocm:nightly
 
 WORKDIR /workspace
 
diff --git a/README.md b/README.md
index c21c830..2dfc4c7 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,6 @@
 # ROCm vLLM 容器化推理项目
 
-基于镜像 `rocm/vllm:rocm7.12.0_gfx120X-all_ubuntu24.04_py3.12_pytorch_2.9.1_vllm_0.16.0` 的 Python + vLLM 推理服务，适配双 AMD R9700 32G GPU。
+基于镜像 `docker.1ms.run/vllm/vllm-openai-rocm:latest` 的 Python + vLLM 推理服务，适配双 AMD R9700 32G GPU。
 
 ## 项目目标
 
@@ -68,7 +68,7 @@
 1. 预拉取基础镜像（与官方文档一致）：
 
 ```bash
-docker pull rocm/vllm:rocm7.12.0_gfx120X-all_ubuntu24.04_py3.12_pytorch_2.9.1_vllm_0.16.0
+docker pull docker.1ms.run/vllm/vllm-openai-rocm:latest
 ```
 
 2. 修改 `config.json` 中的 `models.selected` 与服务参数。
@@ -124,7 +124,7 @@ curl -X POST "http://localhost:8001/v1/chat/completions" \
 
 ## 常见故障排查
 
-- 报错 `model type ... Transformers does not recognize this architecture` 时，先确认已使用当前 Dockerfile 重建镜像（其中会升级 `transformers`），并执行 `docker compose build --no-cache` 后再启动。
+- 报错 `Model architectures ['Qwen3_5MoeForConditionalGeneration'] are not supported for now` 时，在对应 profile 增加 `model_impl=transformers`，并使用当前 Dockerfile 重建镜像后再启动。
 - 报错 `model config (gptq) does not match quantization argument (gptq_marlin)` 时，将该模型配置改为 `dtype=float16` 且 `quantization=gptq`。
 - 报错 `RPC call to sample_tokens timed out` 或出现 `GPU core dump` 时，先下调模型配置为更稳参数：`ctx=32768`、`max_num_seqs=4`、`max_tokens=2048`、`gpu_util=0.90`，并开启 `enforce_eager=true`。
 - 若模型目录存在但仍加载失败，检查挂载路径是否为 `/opt/model:/opt/model:ro`，并确认容器内可见模型文件。
diff --git a/app/model_catalog.py b/app/model_catalog.py
index bdee6c4..eb76872 100644
--- a/app/model_catalog.py
+++ b/app/model_catalog.py
@@ -110,6 +110,7 @@ def resolve_model_profile(
         "served_model_name": profile.get("served_model_name", model_key),
         "dtype": _to_str(profile.get("dtype")),
         "quantization": _to_str(profile.get("quantization")),
+        "model_impl": _to_str(profile.get("model_impl")),
         "reasoning_parser": _to_str(profile.get("reasoning_parser")),
         "max_model_len": _to_int(profile.get("ctx"), 8192),
         "max_num_seqs": _to_int(profile.get("max_num_seqs"), 64),
diff --git a/app/start_openai.py b/app/start_openai.py
index 7dc3f7a..7f1caf3 100644
--- a/app/start_openai.py
+++ b/app/start_openai.py
@@ -25,6 +25,7 @@ def build_command() -> list[str]:
     api_key = runtime["api_key"] or ""
     dtype = str(updates["dtype"] or runtime["dtype"])
     quantization = str(updates["quantization"] or "").strip()
+    model_impl = str(updates["model_impl"] or "").strip()
     reasoning_parser = str(updates["reasoning_parser"] or "").strip()
     revision = runtime["revision"] or ""
     cmd = [
@@ -68,6 +69,8 @@ def build_command() -> list[str]:
         cmd.extend(["--reasoning-parser", reasoning_parser])
     if quantization:
         cmd.extend(["--quantization", quantization])
+    if model_impl:
+        cmd.extend(["--model-impl", model_impl])
     if revision:
         cmd.extend(["--revision", revision])
     if api_key:
diff --git a/config.json b/config.json
index 5d0a2ce..1e35fab 100644
--- a/config.json
+++ b/config.json
@@ -73,6 +73,7 @@
         "local_path": "Qwen3.5-35B-A3B-GPTQ-Int4",
         "dtype": "float16",
         "quantization": "gptq",
+        "model_impl": "transformers",
         "ctx": "32768",
         "trust_remote": true,
         "valid_tp": [2],