diff --git a/Project.toml b/Project.toml
index 5c489a31e..117108f7e 100644
--- a/Project.toml
+++ b/Project.toml
@@ -52,7 +52,7 @@ CEnum = "0.4, 0.5"
 ChainRulesCore = "1"
 EnzymeCore = "0.8"
 ExprTools = "0.1"
-GPUArrays = "11.3.1"
+GPUArrays = "11.3.4"
 GPUCompiler = "1"
 GPUToolbox = "0.1.0, 0.2, 0.3, 1"
 KernelAbstractions = "0.9.2"
diff --git a/src/blas/highlevel.jl b/src/blas/highlevel.jl
index 785b2b778..0f29a35a3 100644
--- a/src/blas/highlevel.jl
+++ b/src/blas/highlevel.jl
@@ -2,7 +2,7 @@ rocblas_size(t::Char, M::ROCVecOrMat) = (size(M, t=='N' ? 1 : 2), size(M, t=='N'
 
 @static if VERSION ≥ v"1.12.0-rc"
     # we need to use the generic wrapper to avoid dispatch to the 2x2or3x3 method
-    using LinearAlgebra: generic_matmatmul_wrapper!, BlasFlag, _symm_hemm_generic!
+    using LinearAlgebra: BlasFlag, _symm_hemm_generic!
 end
 
 #
@@ -165,11 +165,6 @@ end
 #
 
 @static if VERSION ≥ v"1.12.0-rc"
-    function LinearAlgebra.generic_matmatmul_wrapper!(
-        C::StridedROCMatrix{T}, tA::AbstractChar, tB::AbstractChar, A::StridedROCVecOrMat{T}, B::StridedROCVecOrMat{T},
-        alpha::Number, beta::Number, val::LinearAlgebra.BlasFlag.SyrkHerkGemm) where {T<:ROCBLASFloat}
-        LinearAlgebra.generic_matmatmul!(C, tA, tB, A, B, alpha, beta)
-    end
     function LinearAlgebra._symm_hemm_generic!(
         C::StridedROCMatrix{T}, tA::AbstractChar, tB::AbstractChar, A::StridedROCMatrix{T}, B::StridedROCMatrix{T},
         alpha, beta, ::Val{BlasFlag.SYMM}) where {T}