diff --git a/include/flashinfer/attention/handler.cuh b/include/flashinfer/attention/handler.cuh
index 59e383ef..3aa7c115 100644
--- a/include/flashinfer/attention/handler.cuh
+++ b/include/flashinfer/attention/handler.cuh
@@ -180,7 +180,7 @@ cudaError_t BatchDecodeWithPagedKVCacheWorkEstimationDispatched(
     }
     std::tie(max_num_pages_per_batch, new_batch_size) =
         PartitionPagedKVCacheBinarySearchMinNumPagePerBatch(max_grid_size, num_kv_heads, num_pages,
-                                                            128 / page_size);
+                                                            std::max(128 / page_size, 1));
     if (new_batch_size == batch_size && !enable_cuda_graph) {
       // do not use partition-kv kernel for short sequence, when not using CUDAGraph
       split_kv = false;
@@ -564,9 +564,9 @@ cudaError_t PrefillSplitQOKVIndptr(bool& split_kv, uint32_t& split_max_batch_siz
   const uint32_t qo_chunk_size = get_num_rows_per_cta(warp_layout);
 
   // step 2: determine kv_chunk_size
-  std::tie(split_kv, kv_chunk_size, new_batch_size) =
-      PrefillBinarySearchKVChunkSize(max_grid_size, num_kv_heads, packed_qo_len_arr, kv_len_arr,
-                                     qo_chunk_size, /*min_kv_chunk_size=*/(128 / page_size));
+  std::tie(split_kv, kv_chunk_size, new_batch_size) = PrefillBinarySearchKVChunkSize(
+      max_grid_size, num_kv_heads, packed_qo_len_arr, kv_len_arr, qo_chunk_size,
+      /*min_kv_chunk_size=*/std::max((128 / page_size), 1));
 
   // step 3: split qo_indptr and kv_indptr
   total_num_tiles_q = 0;