diff --git a/swift/llm/export/quant.py b/swift/llm/export/quant.py
index 4e598c03..f3e17cde 100644
--- a/swift/llm/export/quant.py
+++ b/swift/llm/export/quant.py
@@ -210,6 +210,7 @@ class QuantEngine(ProcessorMixin):
                 bits=args.quant_bits,
                 dataset=','.join(args.dataset),
                 batch_size=args.quant_batch_size,
+                group_size=args.group_size,
                 block_name_to_quantize=self.get_block_name_to_quantize(self.model, args.model_type))
             gptq_quantizer.serialization_keys.append('block_name_to_quantize')
             logger.info('Start quantizing the model...')