diff --git a/src/rapidata/rapidata_client/benchmark/rapidata_benchmark.py b/src/rapidata/rapidata_client/benchmark/rapidata_benchmark.py
index 2009a68747..6bedce34a3 100644
--- a/src/rapidata/rapidata_client/benchmark/rapidata_benchmark.py
+++ b/src/rapidata/rapidata_client/benchmark/rapidata_benchmark.py
@@ -630,6 +630,7 @@ def create_leaderboard(
         included_tags: list[str] | None = None,
         excluded_tags: list[str] | None = None,
         vote_aggregation: VoteAggregation = VoteAggregation.MAJORITY_VOTE,
+        skip_initial_run: bool = False,
     ) -> RapidataLeaderboard:
         """
         Creates a new leaderboard for the benchmark.
@@ -647,6 +648,7 @@ def create_leaderboard(
             included_tags: Restricts **which of the benchmark's prompts this leaderboard collects matchups for**: only prompts carrying at least one of these tag values are used. When empty or not specified (the default) every prompt is eligible. Note that a non-empty list drops untagged prompts. (default: None)
             excluded_tags: Prompt tag values to skip when collecting matchups. Always wins over ``included_tags`` — a prompt carrying both an included and an excluded tag is skipped. (default: None)
             vote_aggregation: How the responses on a single matchup are aggregated into that matchup's result. :attr:`VoteAggregation.MAJORITY_VOTE` (the default) collapses each matchup to one win for the majority side, ties split 0.5/0.5, so every matchup weighs the same no matter how many responses it collected. :attr:`VoteAggregation.ALL_VOTES` counts every individual response as its own matchup, which lets heavily-answered matchups dominate the standings. Changeable afterwards via :attr:`RapidataLeaderboard.vote_aggregation`.
+            skip_initial_run: Whether to skip the initial run that evaluates the models already in the benchmark against each other. Adding a model afterwards still compares it against the whole existing field, and boosting the leaderboard still works — you just start with no responses collected and therefore no standings. Set this when you want to choose what gets evaluated first instead of paying for a full round. (default: False)
 
         Do not confuse either tag argument with the ``tags`` argument of
         :meth:`RapidataLeaderboard.get_standings`: that filters the standings you read
@@ -693,7 +695,7 @@ def create_leaderboard(
             )
 
             logger.info(
-                "Creating leaderboard %s with instruction %s, show_prompt %s, show_prompt_asset %s, inverse_ranking %s, level_of_detail %s, min_responses_per_matchup %s, audience_id %s, settings %s, included_tags %s, excluded_tags %s, vote_aggregation %s",
+                "Creating leaderboard %s with instruction %s, show_prompt %s, show_prompt_asset %s, inverse_ranking %s, level_of_detail %s, min_responses_per_matchup %s, audience_id %s, settings %s, included_tags %s, excluded_tags %s, vote_aggregation %s, skip_initial_run %s",
                 name,
                 instruction,
                 show_prompt,
@@ -706,6 +708,7 @@ def create_leaderboard(
                 included_tags,
                 excluded_tags,
                 vote_aggregation.name,
+                skip_initial_run,
             )
 
             leaderboard_result = (
@@ -723,6 +726,7 @@ def create_leaderboard(
                         includedTags=included_tags,
                         excludedTags=excluded_tags,
                         voteAggregation=vote_aggregation._to_backend_model(),
+                        skipInitialRun=skip_initial_run,
                         featureFlags=(
                             [setting._to_feature_flag() for setting in settings]
                             if settings
diff --git a/tests/rapidata_client/benchmark/test_leaderboard_skip_initial_run.py b/tests/rapidata_client/benchmark/test_leaderboard_skip_initial_run.py
new file mode 100644
index 0000000000..5890f4671c
--- /dev/null
+++ b/tests/rapidata_client/benchmark/test_leaderboard_skip_initial_run.py
@@ -0,0 +1,55 @@
+"""Tests for ``skip_initial_run`` on leaderboard creation.
+
+The flag suppresses the initial run that evaluates the benchmark's existing
+models against each other. It is create-only: the backend decides it at creation
+time and does not record it on the leaderboard, so there is nothing to read back.
+"""
+
+from __future__ import annotations
+
+from unittest.mock import MagicMock
+
+from rapidata.api_client.models.vote_aggregation import (
+    VoteAggregation as VoteAggregationModel,
+)
+from rapidata.rapidata_client.benchmark.rapidata_benchmark import RapidataBenchmark
+
+
+def _make_benchmark() -> RapidataBenchmark:
+    svc = MagicMock()
+    svc.environment = "rapidata.ai"
+    create = svc.leaderboard.leaderboard_api.leaderboard_post
+    create.return_value.benchmark_id = "bm-1"
+    create.return_value.id = "lb-1"
+    create.return_value.response_budget = 2000
+    create.return_value.min_responses = 3
+    create.return_value.vote_aggregation = VoteAggregationModel.MAJORITYVOTE
+    return RapidataBenchmark("bm", "bm-1", svc)
+
+
+def _sent_payload(benchmark: RapidataBenchmark):
+    create = benchmark._openapi_service.leaderboard.leaderboard_api.leaderboard_post
+    return create.call_args.kwargs["create_leaderboard_endpoint_input"]
+
+
+def test_create_leaderboard_threads_skip_initial_run_to_the_wire() -> None:
+    benchmark = _make_benchmark()
+    benchmark.create_leaderboard(
+        name="lb",
+        instruction="Which is better?",
+        skip_initial_run=True,
+    )
+
+    assert _sent_payload(benchmark).skip_initial_run is True
+
+
+def test_create_leaderboard_runs_initially_by_default() -> None:
+    benchmark = _make_benchmark()
+    benchmark.create_leaderboard(
+        name="lb",
+        instruction="Which is better?",
+    )
+
+    # Explicitly False rather than omitted, so the request never relies on the
+    # server-side default to keep today's behaviour.
+    assert _sent_payload(benchmark).skip_initial_run is False