diff --git a/Project.toml b/Project.toml
index cdc30373..c9f68870 100644
--- a/Project.toml
+++ b/Project.toml
@@ -4,13 +4,12 @@ authors = ["Sungho Shin <sushin@mit.edu>"]
 version = "0.9.5"
 
 [deps]
+Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
 NLPModels = "a4795742-8479-5a88-8948-cc11e1c8c1a6"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 SolverCore = "ff4d7338-4cf1-434d-91df-b86cb86fb843"
 
 [weakdeps]
-AMDGPU = "21141c5a-9bdb-4563-92ae-f87d6854732e"
-CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
 Ipopt = "b6b21f68-93f8-5de0-b562-5493be1d77c9"
 JuMP = "4076af6c-e467-56ae-b986-b466b2749572"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
@@ -20,24 +19,19 @@ MathOptInterface = "b8f27783-ece8-5eb3-8dc8-9495eed66fee"
 NLPModelsIpopt = "f4238b75-b362-5c4c-b852-0801c9a21d71"
 OpenCL = "08131aa3-fb12-5dee-8b74-c09406e224a2"
 SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
-oneAPI = "8f75cd03-7ff8-4ecb-9b8f-daf728133b1b"
 
 [extensions]
-ExaModelsAMDGPU = "AMDGPU"
-ExaModelsCUDA = "CUDA"
 ExaModelsIpopt = ["MathOptInterface", "NLPModelsIpopt"]
 ExaModelsJuMP = "JuMP"
 ExaModelsKernelAbstractions = "KernelAbstractions"
 ExaModelsLinearAlgebra = "LinearAlgebra"
 ExaModelsMOI = "MathOptInterface"
 ExaModelsMadNLP = ["MadNLP", "MathOptInterface"]
-ExaModelsOneAPI = "oneAPI"
 ExaModelsOpenCL = "OpenCL"
 ExaModelsSpecialFunctions = "SpecialFunctions"
 
 [compat]
-AMDGPU = "2"
-CUDA = "5"
+Adapt = "4"
 Ipopt = "1.11"
 JuMP = "1"
 KernelAbstractions = "0.9"
@@ -49,4 +43,3 @@ OpenCL = "0.10"
 SolverCore = "0.3"
 SpecialFunctions = "2"
 julia = "1.9"
-oneAPI = "2"
diff --git a/ext/ExaModelsAMDGPU.jl b/ext/ExaModelsAMDGPU.jl
deleted file mode 100644
index 5f863384..00000000
--- a/ext/ExaModelsAMDGPU.jl
+++ /dev/null
@@ -1,7 +0,0 @@
-module ExaModelsAMDGPU
-
-import ExaModels, AMDGPU
-
-ExaModels.convert_array(v, backend::AMDGPU.ROCBackend) = AMDGPU.ROCArray(v)
-
-end
diff --git a/ext/ExaModelsCUDA.jl b/ext/ExaModelsCUDA.jl
deleted file mode 100644
index 4eeee4cd..00000000
--- a/ext/ExaModelsCUDA.jl
+++ /dev/null
@@ -1,8 +0,0 @@
-module ExaModelsCUDA
-
-import ExaModels: ExaModels, NLPModels
-import CUDA: CUDA, CUDABackend, CuArray
-
-ExaModels.convert_array(v, backend::CUDABackend) = CuArray(v)
-
-end
diff --git a/ext/ExaModelsKernelAbstractions.jl b/ext/ExaModelsKernelAbstractions.jl
index 111f4b75..6b4a2cc0 100644
--- a/ext/ExaModelsKernelAbstractions.jl
+++ b/ext/ExaModelsKernelAbstractions.jl
@@ -3,8 +3,6 @@ module ExaModelsKernelAbstractions
 import ExaModels: ExaModels, NLPModels
 import KernelAbstractions: KernelAbstractions, @kernel, @index, @Const, synchronize, CPU
 
-ExaModels.convert_array(v, backend::CPU) = v
-
 function getitr(gen::UnitRange{Int64})
     return gen
 end
diff --git a/ext/ExaModelsOneAPI.jl b/ext/ExaModelsOneAPI.jl
deleted file mode 100644
index 844ab515..00000000
--- a/ext/ExaModelsOneAPI.jl
+++ /dev/null
@@ -1,19 +0,0 @@
-module ExaModelsOneAPI
-
-import ExaModels, oneAPI
-
-ExaModels.convert_array(v, backend::oneAPI.oneAPIBackend) = oneAPI.oneArray(v)
-
-ExaModels.sort!(array::A; lt = isless) where {A<:oneAPI.oneArray} =
-    copyto!(array, sort!(Array(array); lt = lt))
-
-# below is type piracy
-function Base.findall(f::F, bitarray::A) where {F<:Function,A<:oneAPI.oneArray}
-    a = Array(bitarray)
-    b = findall(f, a)
-    c = similar(bitarray, eltype(b), length(b))
-    return copyto!(c, b)
-end
-Base.findall(bitarray::A) where {A<:oneAPI.oneArray} = Base.findall(identity, bitarray)
-
-end # module
diff --git a/ext/ExaModelsOpenCL.jl b/ext/ExaModelsOpenCL.jl
index f7f3a496..8d24a965 100644
--- a/ext/ExaModelsOpenCL.jl
+++ b/ext/ExaModelsOpenCL.jl
@@ -2,8 +2,6 @@ module ExaModelsOpenCL
 
 import ExaModels, OpenCL
 
-ExaModels.convert_array(v, backend::OpenCL.OpenCLBackend) = OpenCL.CLArray(v)
-
 ExaModels.sort!(array::A; lt = isless) where {A<:OpenCL.CLArray} =
     copyto!(array, sort!(Array(array); lt = lt))
 
diff --git a/src/ExaModels.jl b/src/ExaModels.jl
index f5d0b539..89674485 100644
--- a/src/ExaModels.jl
+++ b/src/ExaModels.jl
@@ -7,6 +7,7 @@ For more information, please visit https://github.com/exanauts/ExaModels.jl
 """
 module ExaModels
 
+import Adapt: adapt
 import NLPModels:
     NLPModels,
     obj,
diff --git a/src/templates.jl b/src/templates.jl
index 524023bf..a3dd5ce2 100644
--- a/src/templates.jl
+++ b/src/templates.jl
@@ -1,5 +1,6 @@
-# A template for convert_array. This is extended in extension packages for each device architecture.
+# Convert array to the appropriate device using Adapt.jl.
 convert_array(v, ::Nothing) = v
+convert_array(v, backend) = adapt(backend, v)
 
 # to avoid type privacy
 sort!(array; kwargs...) = Base.sort!(array; kwargs...)