diff --git a/README.md b/README.md
index 80c8ef57..4341c8f2 100755
--- a/README.md
+++ b/README.md
@@ -40,6 +40,12 @@ conda env create -n sklearn -f envs/conda-env-sklearn.yml
 conda env create -n rapids --solver=libmamba -f envs/conda-env-rapids.yml
 ```
 
+GPU benchmarks using `sklearnex`'s `device: gpu` rely on scikit-learn's Array API support (`array_api_dispatch`), which requires `SCIPY_ARRAY_API=1` to be set in the environment before running:
+
+```bash
+export SCIPY_ARRAY_API=1
+```
+
 ## 🚀 How To Use Scikit-learn_bench
 
 ### Benchmarks Runner
diff --git a/configs/BENCH-CONFIG-SPEC.md b/configs/BENCH-CONFIG-SPEC.md
index c34f4743..7e23e548 100644
--- a/configs/BENCH-CONFIG-SPEC.md
+++ b/configs/BENCH-CONFIG-SPEC.md
@@ -98,7 +98,7 @@ Configs have the three highest parameter keys:
 | `data`:`distributed_split` | None | None, `rank_based` | Split type used to distribute data between machines in distributed algorithm. `None` type means usage of all data without split on all machines. `rank_based` type splits the data equally between machines with split sequence based on rank id from MPI. |
 |<h3>Algorithm parameters</h3>||||
 | `algorithm`:`library` | None |  | Python module containing measured entity (class or function). |
-| `algorithm`:`device` | `default` | `default`, `cpu`, `gpu` | Device selected for computation. |
+| `algorithm`:`device` | `default` | `default`, `cpu`, `gpu` | Device selected for computation. `sklearnex`+`gpu` cases enable sklearn's `array_api_dispatch` and use `dpnp` data by default (see `sklearn_context` below); requires `SCIPY_ARRAY_API=1` to be set in the environment. |
 
 ## Benchmark-Specific Parameters
 
diff --git a/configs/common/sklearn.json b/configs/common/sklearn.json
index 6b988ba7..0ac26ef8 100644
--- a/configs/common/sklearn.json
+++ b/configs/common/sklearn.json
@@ -6,12 +6,18 @@
                 { "library": "sklearnex", "device": "cpu" }
             ]
         },
-        "sklearn-ex[cpu,gpu] implementations": {
-            "algorithm": [
-                { "library": "sklearn", "device": "cpu" },
-                { "library": "sklearnex", "device": ["cpu", "gpu"] }
-            ]
-        },
+        "sklearn-ex[cpu,gpu] implementations": [
+            { "algorithm": { "library": "sklearn", "device": "cpu" } },
+            { "algorithm": { "library": "sklearnex", "device": "cpu" } },
+            {
+                "algorithm": {
+                    "library": "sklearnex",
+                    "device": "gpu",
+                    "sklearn_context": { "array_api_dispatch": true }
+                },
+                "data": { "format": "dpnp" }
+            }
+        ],
         "sklearnex spmd implementation": {
             "algorithm": {
                 "library": "sklearnex.spmd",