diff --git a/nncf/quantization/algorithms/weight_compression/weight_lowering.py b/nncf/quantization/algorithms/weight_compression/weight_lowering.py
index d5828910083..36a55721693 100644
--- a/nncf/quantization/algorithms/weight_compression/weight_lowering.py
+++ b/nncf/quantization/algorithms/weight_compression/weight_lowering.py
@@ -169,7 +169,7 @@ def calculate_signed_scale(weight: Tensor, reduction_axes: ReductionAxes, num_bi
     w_abs_min = fns.abs(fns.min(weight, axis=reduction_axes, keepdims=True))
     w_max = fns.max(weight, axis=reduction_axes, keepdims=True)
 
-    scale = fns.where(w_abs_min > w_max, w_abs_min, -w_max)
+    scale = fns.where(w_abs_min >= w_max, w_abs_min, -w_max)
     scale /= level_high
 
     eps = fns.finfo(scale).eps