diff --git a/src/voxcpm/model/voxcpm.py b/src/voxcpm/model/voxcpm.py
index 55c96e9..ed3fe18 100644
--- a/src/voxcpm/model/voxcpm.py
+++ b/src/voxcpm/model/voxcpm.py
@@ -47,6 +47,16 @@ from ..modules.minicpm4 import MiniCPM4Config, MiniCPMModel
 from .utils import get_dtype, mask_multichar_chinese_tokens, resolve_runtime_device
 
 
+# Ref: https://github.com/OpenBMB/VoxCPM/issues/256#issuecomment-4234809321
+# Explicitly close partially-consumed generators so inference_mode cleanup
+# does not get deferred to Python's GC/finalizer path.
+def _next_and_close(gen):
+    try:
+        return next(gen)
+    finally:
+        gen.close()
+
+
 class VoxCPMEncoderConfig(BaseModel):
     hidden_dim: int = 1024
     ffn_dim: int = 4096
@@ -335,7 +345,7 @@ class VoxCPMModel(nn.Module):
         return get_dtype(self.config.dtype)
 
     def generate(self, *args, **kwargs) -> torch.Tensor:
-        return next(self._generate(*args, streaming=False, **kwargs))
+        return _next_and_close(self._generate(*args, streaming=False, **kwargs))
 
     def generate_streaming(self, *args, **kwargs) -> Generator[torch.Tensor, None, None]:
         return self._generate(*args, streaming=True, **kwargs)
@@ -461,7 +471,7 @@ class VoxCPMModel(nn.Module):
                     yield decode_audio
                 break
             else:
-                latent_pred, pred_audio_feat = next(inference_result)
+                latent_pred, pred_audio_feat = _next_and_close(inference_result)
                 if retry_badcase:
                     if pred_audio_feat.shape[0] >= target_text_length * retry_badcase_ratio_threshold:
                         print(
@@ -569,7 +579,7 @@ class VoxCPMModel(nn.Module):
         return merged_cache
 
     def generate_with_prompt_cache(self, *args, **kwargs) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
-        return next(self._generate_with_prompt_cache(*args, streaming=False, **kwargs))
+        return _next_and_close(self._generate_with_prompt_cache(*args, streaming=False, **kwargs))
 
     def generate_with_prompt_cache_streaming(
         self, *args, **kwargs
@@ -688,7 +698,7 @@ class VoxCPMModel(nn.Module):
                     yield (decode_audio, target_text_token, pred_audio_feat)
                 break
             else:
-                latent_pred, pred_audio_feat = next(inference_result)
+                latent_pred, pred_audio_feat = _next_and_close(inference_result)
                 if retry_badcase:
                     if pred_audio_feat.shape[0] >= target_text_length * retry_badcase_ratio_threshold:
                         print(
@@ -711,7 +721,7 @@ class VoxCPMModel(nn.Module):
             yield (decode_audio, target_text_token, pred_audio_feat)
 
     def inference(self, *args, **kwargs) -> Tuple[torch.Tensor, torch.Tensor]:
-        return next(self._inference(*args, streaming=False, **kwargs))
+        return _next_and_close(self._inference(*args, streaming=False, **kwargs))
 
     def inference_streaming(self, *args, **kwargs) -> Generator[Tuple[torch.Tensor, List[torch.Tensor]], None, None]:
         return self._inference(*args, streaming=True, **kwargs)
diff --git a/src/voxcpm/model/voxcpm2.py b/src/voxcpm/model/voxcpm2.py
index 3ef5592..f523872 100644
--- a/src/voxcpm/model/voxcpm2.py
+++ b/src/voxcpm/model/voxcpm2.py
@@ -48,6 +48,16 @@ from ..modules.minicpm4 import MiniCPM4Config, MiniCPMModel
 from .utils import get_dtype, mask_multichar_chinese_tokens, resolve_runtime_device
 
 
+# Ref: https://github.com/OpenBMB/VoxCPM/issues/256#issuecomment-4234809321
+# Explicitly close partially-consumed generators so inference_mode cleanup
+# does not get deferred to Python's GC/finalizer path.
+def _next_and_close(gen):
+    try:
+        return next(gen)
+    finally:
+        gen.close()
+
+
 # A simple function to trim audio silence using VAD, not used default
 def _trim_audio_silence_vad(audio: torch.Tensor, sample_rate: int, max_silence_ms: float = 200.0, top_db: float = 35.0) -> torch.Tensor:
     if audio.numel() == 0:
@@ -441,7 +451,7 @@ class VoxCPM2Model(nn.Module):
         return tokens, feats, t_mask, a_mask
 
     def generate(self, *args, **kwargs) -> torch.Tensor:
-        return next(self._generate(*args, streaming=False, **kwargs))
+        return _next_and_close(self._generate(*args, streaming=False, **kwargs))
 
     def generate_streaming(self, *args, **kwargs) -> Generator[torch.Tensor, None, None]:
         return self._generate(*args, streaming=True, **kwargs)
@@ -641,7 +651,7 @@ class VoxCPM2Model(nn.Module):
                     yield decode_audio
                 break
             else:
-                latent_pred, pred_audio_feat, context_len = next(inference_result)
+                latent_pred, pred_audio_feat, context_len = _next_and_close(inference_result)
                 if retry_badcase:
                     if pred_audio_feat.shape[0] >= target_text_length * retry_badcase_ratio_threshold:
                         print(
@@ -759,7 +769,7 @@ class VoxCPM2Model(nn.Module):
         return merged
 
     def generate_with_prompt_cache(self, *args, **kwargs) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
-        return next(self._generate_with_prompt_cache(*args, streaming=False, **kwargs))
+        return _next_and_close(self._generate_with_prompt_cache(*args, streaming=False, **kwargs))
 
     def generate_with_prompt_cache_streaming(
         self, *args, **kwargs
@@ -928,7 +938,7 @@ class VoxCPM2Model(nn.Module):
                     yield (decode_audio, target_text_token, pred_audio_feat)
                 break
             else:
-                latent_pred, pred_audio_feat, context_len = next(inference_result)
+                latent_pred, pred_audio_feat, context_len = _next_and_close(inference_result)
                 if retry_badcase:
                     if pred_audio_feat.shape[0] >= target_text_length * retry_badcase_ratio_threshold:
                         print(
@@ -951,7 +961,7 @@ class VoxCPM2Model(nn.Module):
             yield (decode_audio, target_text_token, pred_audio_feat)
 
     def inference(self, *args, **kwargs) -> Tuple[torch.Tensor, torch.Tensor]:
-        feat_pred, generated_feat, _ = next(self._inference(*args, streaming=False, **kwargs))
+        feat_pred, generated_feat, _ = _next_and_close(self._inference(*args, streaming=False, **kwargs))
         return feat_pred, generated_feat
 
     def inference_streaming(self, *args, **kwargs) -> Generator[Tuple[torch.Tensor, List[torch.Tensor]], None, None]: