diff --git a/ci.jsonnet b/ci.jsonnet
index de2077d15e..913dfebc61 100644
--- a/ci.jsonnet
+++ b/ci.jsonnet
@@ -5,7 +5,7 @@
 (import "ci/python-gate.libsonnet") +
 (import "ci/python-bench.libsonnet") +
 {
-    overlay: "12367561e6a2b54df8d3b1bd400e431de01eef0f",
+    overlay: "acd356580b18f9d67b0213c12da04076e6a4d493",
     specVersion: "8",
     // Until buildbot issues around CI tiers are resolved, we cannot use them
     // tierConfig: self.tierConfig,
@@ -345,6 +345,28 @@
             "raw_results.json",
         ],
     }),
+    // Existing GraalPy post-merge benchmark jobs provide master history for
+    // PR-bench. The report uses that history as a best-effort baseline and
+    // does not run separate baseline benchmarking.
+    local pr_bench_graalpy = task_spec({
+        unicorn_pull_request_benchmarking:: {
+            name: "graalpy",
+            benchmarks: [
+                "meso.nbody3",
+                "macro.c-pydantic-validate",
+                "macro.c-pydantic-serialize-json",
+                "micro-small.attribute-bool",
+            ],
+            metrics: ["time"],
+            baseline_benchmarking: false,
+            _extra_unicorn_args: [
+                // Use JSON to preserve the string types expected by Bench Server;
+                // --config-key would coerce numeric-looking values such as "0".
+                "--config-json",
+                "{\"dataserver\":{\"fetch\":{\"where\":{\"bench-suite-version\":{\"enum\":[\"unknown\"],\"allow-absent\":true},\"platform.jdk-major-version\":{\"enum\":[\"0\"],\"allow-absent\":true}}}}}",
+            ],
+        },
+    }),
 
     // -----------------------------------------------------------------------------------------------------------------
     // benchmarks
@@ -352,7 +374,7 @@
     // [info]: when adding a benchmark, the key in the `bench_task_dict` is taken as the name of the benchmark if it is
     // not specified as the first arg to `bench_task`.
     local bench_task_dict = {
-        [bench]: bench_task(bench) + platform_spec(no_jobs) + bench_variants({
+        "micro": bench_task("micro") + platform_spec(no_jobs) + bench_variants({
             "vm_name:graalvm_ee_default"                                : {"linux:amd64:jdk-latest" : post_merge + t("08:00:00") + need_pgo},
             "vm_name:graalpython_enterprise"                            : {"linux:amd64:jdk-latest" : daily      + t("08:00:00"),
                 "job_type:checkup"                                      : {"linux:amd64:jdk-latest" : on_demand  + t("08:00:00")}
@@ -361,10 +383,18 @@
             "vm_name:cpython"                                           : {"linux:amd64:jdk-latest" : monthly    + t("04:00:00")},
             "vm_name:pypy"                                              : {"linux:amd64:jdk-latest" : on_demand    + t("04:00:00")},
         }),
-        for bench in ["micro", "meso"]
+        "meso": bench_task("meso") + platform_spec(no_jobs) + bench_variants({
+            "vm_name:graalvm_ee_default"                                : {"linux:amd64:jdk-latest" : post_merge + t("08:00:00") + need_pgo + pr_bench_graalpy},
+            "vm_name:graalpython_enterprise"                            : {"linux:amd64:jdk-latest" : daily      + t("08:00:00"),
+                "job_type:checkup"                                      : {"linux:amd64:jdk-latest" : on_demand  + t("08:00:00")}
+            },
+            "vm_name:graalpython_enterprise_multi"                      : {"linux:amd64:jdk-latest" : weekly     + t("08:00:00")},
+            "vm_name:cpython"                                           : {"linux:amd64:jdk-latest" : monthly    + t("04:00:00")},
+            "vm_name:pypy"                                              : {"linux:amd64:jdk-latest" : on_demand    + t("04:00:00")},
+        }),
     } + {
         "macro": bench_task("macro") + internet_access_env + platform_spec(no_jobs) + bench_variants({
-            "vm_name:graalvm_ee_default"                                : {"linux:amd64:jdk-latest" : post_merge + t("08:00:00") + need_pgo},
+            "vm_name:graalvm_ee_default"                                : {"linux:amd64:jdk-latest" : post_merge + t("08:00:00") + need_pgo + pr_bench_graalpy},
             "vm_name:graalpython_enterprise"                            : {"linux:amd64:jdk-latest" : daily      + t("08:00:00"),
                 "job_type:checkup"                                      : {"linux:amd64:jdk-latest" : on_demand  + t("08:00:00")}
             },
@@ -390,13 +420,18 @@
     } + {
         // "small" benchmarks have their argument set such that they run in a resonable
         // time in the interpreter and they are used for interpreter benchmarking
-        [bench]: bench_task(bench) + platform_spec(no_jobs) + bench_variants({
+        "micro_small": bench_task("micro_small") + platform_spec(no_jobs) + bench_variants({
+            "vm_name:graalvm_ee_default_interpreter"                    : {"linux:amd64:jdk-latest" : post_merge + t("02:00:00") + need_pgo + pr_bench_graalpy},
+            "vm_name:graalvm_ee_default_interpreter_uncached"           : {"linux:amd64:jdk-latest" : daily      + t("02:00:00") + need_pgo},
+            "vm_name:graalpython_enterprise_interpreter"                : {"linux:amd64:jdk-latest" : weekly     + t("02:00:00")},
+            "vm_name:cpython"                                           : {"linux:amd64:jdk-latest" : weekly     + t("02:00:00")},
+        }),
+        "meso_small": bench_task("meso_small") + platform_spec(no_jobs) + bench_variants({
             "vm_name:graalvm_ee_default_interpreter"                    : {"linux:amd64:jdk-latest" : daily      + t("02:00:00") + need_pgo},
             "vm_name:graalvm_ee_default_interpreter_uncached"           : {"linux:amd64:jdk-latest" : daily      + t("02:00:00") + need_pgo},
             "vm_name:graalpython_enterprise_interpreter"                : {"linux:amd64:jdk-latest" : weekly     + t("02:00:00")},
             "vm_name:cpython"                                           : {"linux:amd64:jdk-latest" : weekly     + t("02:00:00")},
         }),
-        for bench in ["micro_small", "meso_small"]
     } + {
         // benchmarks executed via Java embedding driver
         [bench]: bench_task(bench) + platform_spec(no_jobs) + bench_variants({