diff --git a/Project.toml b/Project.toml
index 02a02f68..50de5fe1 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "ApproximateGPs"
 uuid = "298c2ebc-0411-48ad-af38-99e88101b606"
 authors = ["JuliaGaussianProcesses Team"]
-version = "0.4.5"
+version = "0.4.6"
 
 [deps]
 AbstractGPs = "99985d1d-32ba-4be9-9821-2ec096f28918"
@@ -23,7 +23,7 @@ StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [compat]
-AbstractGPs = "0.3, 0.4, 0.5"
+AbstractGPs = "0.5.17"
 ChainRulesCore = "1.7"
 Distributions = "0.25"
 FastGaussQuadrature = "0.4, 0.5"
diff --git a/src/API.jl b/src/API.jl
index cc710869..a8893a47 100644
--- a/src/API.jl
+++ b/src/API.jl
@@ -1,15 +1,7 @@
 module API
 
-export approx_lml  # TODO move to AbstractGPs, see https://github.com/JuliaGaussianProcesses/AbstractGPs.jl/issues/221
+export approx_lml  # this was moved into AbstractGPs.approx_log_evidence, see https://github.com/JuliaGaussianProcesses/AbstractGPs.jl/pull/361
 
-"""
-    approx_lml(approx::<Approximation>, lfx::LatentFiniteGP, ys)
-
-Compute an approximation to the log of the marginal likelihood (also known as
-"evidence") under the given `approx` to the posterior. This approximation can be used to optimise the hyperparameters of `lfx`.
-
-This should become part of the AbstractGPs API (see JuliaGaussianProcesses/AbstractGPs.jl#221).
-"""
-function approx_lml end
+@deprecate approx_lml(approx, lfx, ys) approx_log_evidence(approx, lfx, ys)
 
 end
diff --git a/src/LaplaceApproximationModule.jl b/src/LaplaceApproximationModule.jl
index e596a72d..eb93fef9 100644
--- a/src/LaplaceApproximationModule.jl
+++ b/src/LaplaceApproximationModule.jl
@@ -48,14 +48,14 @@ function AbstractGPs.posterior(la::LaplaceApproximation, lfx::LatentFiniteGP, ys
 end
 
 """
-    approx_lml(la::LaplaceApproximation, lfx::LatentFiniteGP, ys)
+    approx_log_evidence(la::LaplaceApproximation, lfx::LatentFiniteGP, ys)
 
 Compute an approximation to the log of the marginal likelihood (also known as
 "evidence"), which can be used to optimise the hyperparameters of `lfx`.
 
 This should become part of the AbstractGPs API (see JuliaGaussianProcesses/AbstractGPs.jl#221).
 """
-function API.approx_lml(la::LaplaceApproximation, lfx::LatentFiniteGP, ys)
+function AbstractGPs.approx_log_evidence(la::LaplaceApproximation, lfx::LatentFiniteGP, ys)
     return laplace_lml(lfx, ys; la.newton_kwargs...)
 end
 
diff --git a/src/SparseVariationalApproximationModule.jl b/src/SparseVariationalApproximationModule.jl
index 1f80938e..1dbbfaf1 100644
--- a/src/SparseVariationalApproximationModule.jl
+++ b/src/SparseVariationalApproximationModule.jl
@@ -273,10 +273,10 @@ inducing_points(f::ApproxPosteriorGP{<:SparseVariationalApproximation}) = f.appr
 # elbo
 #
 
-function API.approx_lml(
+function AbstractGPs.approx_log_evidence(
     sva::SparseVariationalApproximation, l_fx::Union{FiniteGP,LatentFiniteGP}, ys; kwargs...
 )
-    return AbstractGPs.elbo(sva, l_fx, ys; kwargs...)
+    return elbo(sva, l_fx, ys; kwargs...)
 end
 
 """
diff --git a/src/TestUtils.jl b/src/TestUtils.jl
index 7836e974..cf0689b6 100644
--- a/src/TestUtils.jl
+++ b/src/TestUtils.jl
@@ -36,7 +36,7 @@ function build_latent_gp(theta)
 end
 
 """
-    test_approx_lml(approx)
+    test_approx_log_evidence(approx)
 
 Test whether in the conjugate case `approx_lml(approx, LatentGP(f,
 GaussianLikelihood(), jitter)(x), y)` gives approximately the same answer as
@@ -51,7 +51,7 @@ the log marginal likelihood in exact GP regression.
 !!! todo
     Also test gradients (for hyperparameter optimization).
 """
-function test_approx_lml end
+function test_approx_log_evidence end
 
 """
     test_approximation_predictions(approx)
diff --git a/test/LaplaceApproximationModule.jl b/test/LaplaceApproximationModule.jl
index 547b60f3..f4b0e280 100644
--- a/test/LaplaceApproximationModule.jl
+++ b/test/LaplaceApproximationModule.jl
@@ -33,14 +33,14 @@
     end
 
     @testset "gradients" begin
-        @testset "approx_lml" begin
+        @testset "approx_log_evidence" begin
             X, Y = generate_data()
 
             Random.seed!(123)
             theta0 = rand(2)
             function objective(theta)
                 lf = build_latent_gp(theta)
-                lml = approx_lml(LaplaceApproximation(), lf(X), Y)
+                lml = approx_log_evidence(LaplaceApproximation(), lf(X), Y)
                 return -lml
             end
             fd_grad = only(FiniteDifferences.grad(central_fdm(5, 1), objective, theta0))
@@ -145,7 +145,7 @@
         @testset "reference optimum" begin
             function objective(theta)
                 lf = build_latent_gp(theta)
-                return -approx_lml(LaplaceApproximation(), lf(X), Y)
+                return -approx_log_evidence(LaplaceApproximation(), lf(X), Y)
             end
 
             @testset "NelderMead" begin
diff --git a/test/Project.toml b/test/Project.toml
index 58b2bd88..208d19d6 100644
--- a/test/Project.toml
+++ b/test/Project.toml
@@ -16,7 +16,7 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [compat]
-AbstractGPs = "0.4, 0.5"
+AbstractGPs = "0.5.17"
 ApproximateGPs = "0.4"
 ChainRulesCore = "1"
 ChainRulesTestUtils = "1.2.3"