diff --git a/mlx/types/fp16.h b/mlx/types/fp16.h
index 58e1bc088..c174afebc 100644
--- a/mlx/types/fp16.h
+++ b/mlx/types/fp16.h
@@ -68,7 +68,7 @@ struct _MLX_Float16 {
       inf_scale.u = uint32_t(0x77800000);
       zero_scale.u = uint32_t(0x08800000);
 
-      // Combine with magic and let addition do rouding
+      // Combine with magic and let addition do rounding
       magic_bits.u = x_expo_32;
       magic_bits.f += (std::abs(x) * inf_scale.f) * zero_scale.f;