diff --git a/vllm/v1/core/kv_cache_manager.py b/vllm/v1/core/kv_cache_manager.py
index 0587488d5f67d..59aa38029a69e 100644
--- a/vllm/v1/core/kv_cache_manager.py
+++ b/vllm/v1/core/kv_cache_manager.py
@@ -203,8 +203,8 @@ def allocate_slots(
         # If a computed block of a request is an eviction candidate (in the
         # free queue and ref_cnt == 0), it cannot be counted as a free block
         # when allocating this request.
-        num_evictable_computed_blocks = len(
-            [blk for blk in computed_blocks if blk.ref_cnt == 0])
+        num_evictable_computed_blocks = sum(
+            blk for blk in computed_blocks if blk.ref_cnt == 0)
 
         num_required_blocks = cdiv(num_tokens, self.block_size)
         if (num_required_blocks > self.free_block_queue.num_free_blocks -