diff --git a/include/libvirt/libvirt-host.h b/include/libvirt/libvirt-host.h
index 91214ea21b..19043235b2 100644
--- a/include/libvirt/libvirt-host.h
+++ b/include/libvirt/libvirt-host.h
@@ -974,6 +974,9 @@ typedef enum {
     VIR_CONNECT_BASELINE_CPU_EXPAND_FEATURES  = (1 << 0),
     /* filter out non-migratable features (Since: 1.2.14) */
     VIR_CONNECT_BASELINE_CPU_MIGRATABLE = (1 << 1),
+    /* when computing a baseline from several CPUs, do not make the result
+     * dependent on the current host (Since: 11.5.0) */
+    VIR_CONNECT_BASELINE_CPU_IGNORE_HOST = (1 << 2),
 } virConnectBaselineCPUFlags;
 
 char *virConnectBaselineCPU(virConnectPtr conn,
diff --git a/src/libvirt-host.c b/src/libvirt-host.c
index 8d2107fd62..da75f5f30b 100644
--- a/src/libvirt-host.c
+++ b/src/libvirt-host.c
@@ -1305,7 +1305,8 @@ virConnectBaselineCPU(virConnectPtr conn,
  * it on another host results in an undefined behavior as the computed CPU
  * model is influenced by the hypervisor (the result may use an unexpected CPU
  * model or some features may disabled even though they are supported on all
- * input CPUs).
+ * input CPUs). The undefined behavior can be avoided using
+ * VIR_CONNECT_BASELINE_CPU_IGNORE_HOST flag (see below).
  *
  * This is different from virConnectBaselineCPU() which doesn't consider any
  * hypervisor abilities when computing the best CPU.
@@ -1333,6 +1334,11 @@ virConnectBaselineCPU(virConnectPtr conn,
  * If @flags includes VIR_CONNECT_BASELINE_CPU_MIGRATABLE, the resulting
  * CPU will not include features that block migration.
  *
+ * If @flags contains VIR_CONNECT_BASELINE_CPU_IGNORE_HOST and @ncpus > 1, the
+ * API will not consider hypervisor abilities when computing the baseline. With
+ * this flag baseline can be safely computed on any host (even those not
+ * described in @xmlCPUs).
+ *
  * Returns XML description of the computed CPU (caller frees) or NULL on error.
  *
  * Since: 4.4.0