diff --git a/benchmarks/commit0/run_infer.py b/benchmarks/commit0/run_infer.py
index 36af180a5..90e964911 100644
--- a/benchmarks/commit0/run_infer.py
+++ b/benchmarks/commit0/run_infer.py
@@ -24,7 +24,7 @@
     construct_eval_output_dir,
     get_default_on_result_writer,
 )
-from benchmarks.utils.image_utils import create_docker_workspace, image_exists
+from benchmarks.utils.image_utils import create_docker_workspace, remote_image_exists
 from benchmarks.utils.llm_config import load_llm_config
 from benchmarks.utils.models import (
     EvalInstance,
@@ -213,7 +213,7 @@ def prepare_workspace(
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-{custom_tag}{suffix}"
             )
 
-            if not image_exists(agent_server_image):
+            if not remote_image_exists(agent_server_image):
                 raise RuntimeError(
                     f"Agent server image {agent_server_image} does not exist in container registry. "
                     "Run 'benchmarks/commit0/build_images.py --push' to build and push it first."
diff --git a/benchmarks/gaia/run_infer.py b/benchmarks/gaia/run_infer.py
index fa0317a8e..36e472cbf 100644
--- a/benchmarks/gaia/run_infer.py
+++ b/benchmarks/gaia/run_infer.py
@@ -27,7 +27,7 @@
     get_default_on_result_writer,
 )
 from benchmarks.utils.fake_user_response import run_conversation_with_fake_user_response
-from benchmarks.utils.image_utils import create_docker_workspace, image_exists
+from benchmarks.utils.image_utils import create_docker_workspace, remote_image_exists
 from benchmarks.utils.llm_config import load_llm_config
 from benchmarks.utils.models import EvalInstance, EvalMetadata, EvalOutput
 from benchmarks.utils.version import SDK_SHORT_SHA
@@ -182,7 +182,7 @@ def prepare_workspace(
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-gaia-binary"
             )
 
-            if not image_exists(agent_server_image):
+            if not remote_image_exists(agent_server_image):
                 raise RuntimeError(
                     f"Agent server image {agent_server_image} does not exist in container registry. "
                     f"Run 'benchmarks/gaia/build_images.py --push' to build and push it first."
diff --git a/benchmarks/multiswebench/run_infer.py b/benchmarks/multiswebench/run_infer.py
index ec5137a2d..833fcdcc4 100644
--- a/benchmarks/multiswebench/run_infer.py
+++ b/benchmarks/multiswebench/run_infer.py
@@ -26,7 +26,7 @@
     get_default_on_result_writer,
 )
 from benchmarks.utils.fake_user_response import run_conversation_with_fake_user_response
-from benchmarks.utils.image_utils import image_exists
+from benchmarks.utils.image_utils import remote_image_exists
 from benchmarks.utils.llm_config import load_llm_config
 from benchmarks.utils.models import (
     EvalInstance,
@@ -234,7 +234,7 @@ def prepare_workspace(
             agent_server_image = (
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-{custom_tag}{suffix}"
             )
-            if not image_exists(agent_server_image):
+            if not remote_image_exists(agent_server_image):
                 raise RuntimeError(
                     f"Agent server image {agent_server_image} does not exist in container registry, "
                     "make sure to build, push it, and make it public accessible before using remote workspace."
diff --git a/benchmarks/swebench/build_images.py b/benchmarks/swebench/build_images.py
index cae96b871..5ace5419b 100644
--- a/benchmarks/swebench/build_images.py
+++ b/benchmarks/swebench/build_images.py
@@ -22,7 +22,7 @@
     run_docker_build_layer,
 )
 from benchmarks.utils.dataset import get_dataset
-from benchmarks.utils.image_utils import image_exists
+from benchmarks.utils.image_utils import remote_image_exists
 from openhands.sdk import get_logger
 
 
@@ -94,7 +94,7 @@ def wrap_image(agent_image: str, push: bool = False) -> BuildOutput:
     For pushes, verify the base tag exists in the registry. For local builds,
     assume the tag is available locally or resolvable by Docker during buildx.
     """
-    if push and not image_exists(agent_image):
+    if push and not remote_image_exists(agent_image):
         return BuildOutput(
             base_image=agent_image,
             tags=[],
diff --git a/benchmarks/swebench/run_infer.py b/benchmarks/swebench/run_infer.py
index 231de93e9..5021db16e 100644
--- a/benchmarks/swebench/run_infer.py
+++ b/benchmarks/swebench/run_infer.py
@@ -26,7 +26,7 @@
     get_default_on_result_writer,
 )
 from benchmarks.utils.fake_user_response import run_conversation_with_fake_user_response
-from benchmarks.utils.image_utils import image_exists
+from benchmarks.utils.image_utils import remote_image_exists
 from benchmarks.utils.llm_config import load_llm_config
 from benchmarks.utils.models import (
     EvalInstance,
@@ -195,7 +195,7 @@ def prepare_workspace(
             agent_server_image = (
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-{custom_tag}{suffix}"
             )
-            if not image_exists(agent_server_image):
+            if not remote_image_exists(agent_server_image):
                 raise RuntimeError(
                     f"Agent server image {agent_server_image} does not exist in container registry, "
                     "make sure to build, push it, and make it public accessible before using remote workspace."
diff --git a/benchmarks/swebenchmultimodal/run_infer.py b/benchmarks/swebenchmultimodal/run_infer.py
index 6a1b533a7..fdf5382ad 100644
--- a/benchmarks/swebenchmultimodal/run_infer.py
+++ b/benchmarks/swebenchmultimodal/run_infer.py
@@ -24,7 +24,7 @@
     get_default_on_result_writer,
 )
 from benchmarks.utils.fake_user_response import run_conversation_with_fake_user_response
-from benchmarks.utils.image_utils import image_exists
+from benchmarks.utils.image_utils import remote_image_exists
 from benchmarks.utils.llm_config import load_llm_config
 from benchmarks.utils.models import (
     EvalInstance,
@@ -187,7 +187,7 @@ def prepare_workspace(
             agent_server_image = (
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-{custom_tag}{suffix}"
             )
-            if not image_exists(agent_server_image):
+            if not remote_image_exists(agent_server_image):
                 raise RuntimeError(
                     f"Agent server image {agent_server_image} does not exist in container registry, "
                     "make sure to build, push it, and make it public accessible before using remote workspace."
diff --git a/benchmarks/swefficiency/run_infer.py b/benchmarks/swefficiency/run_infer.py
index cf5a902d2..f559d7f6e 100644
--- a/benchmarks/swefficiency/run_infer.py
+++ b/benchmarks/swefficiency/run_infer.py
@@ -21,7 +21,7 @@
     get_default_on_result_writer,
 )
 from benchmarks.utils.fake_user_response import run_conversation_with_fake_user_response
-from benchmarks.utils.image_utils import image_exists
+from benchmarks.utils.image_utils import remote_image_exists
 from benchmarks.utils.models import (
     EvalInstance,
     EvalMetadata,
@@ -248,7 +248,7 @@ def prepare_workspace(
             remote_agent_image = (
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-{custom_tag}{suffix}"
             )
-            if not image_exists(remote_agent_image):
+            if not remote_image_exists(remote_agent_image):
                 raise RuntimeError(
                     f"Agent server image {remote_agent_image} does not exist in container registry, "
                     "make sure to build, push it, and make it public accessible before using remote workspace."
diff --git a/benchmarks/swtbench/build_eval_env_images.py b/benchmarks/swtbench/build_eval_env_images.py
index fde30ed9c..2f0ea9862 100644
--- a/benchmarks/swtbench/build_eval_env_images.py
+++ b/benchmarks/swtbench/build_eval_env_images.py
@@ -12,7 +12,7 @@
 from benchmarks.swtbench.config import EVAL_DEFAULTS
 from benchmarks.swtbench.image_utils import ensure_swt_bench_repo
 from benchmarks.utils.dataset import get_dataset
-from benchmarks.utils.image_utils import image_exists as remote_image_exists
+from benchmarks.utils.image_utils import remote_image_exists
 from openhands.sdk import get_logger
 
 
diff --git a/benchmarks/swtbench/run_infer.py b/benchmarks/swtbench/run_infer.py
index 36b78f265..132949427 100644
--- a/benchmarks/swtbench/run_infer.py
+++ b/benchmarks/swtbench/run_infer.py
@@ -18,7 +18,7 @@
     get_default_on_result_writer,
 )
 from benchmarks.utils.fake_user_response import run_conversation_with_fake_user_response
-from benchmarks.utils.image_utils import create_docker_workspace, image_exists
+from benchmarks.utils.image_utils import create_docker_workspace, remote_image_exists
 from benchmarks.utils.llm_config import load_llm_config
 from benchmarks.utils.models import (
     EvalInstance,
@@ -188,7 +188,7 @@ def prepare_workspace(
             agent_server_image = (
                 f"{EVAL_AGENT_SERVER_IMAGE}:{sdk_short_sha}-{custom_tag}{suffix}"
             )
-            if not image_exists(agent_server_image):
+            if not remote_image_exists(agent_server_image):
                 raise RuntimeError(
                     f"Agent server image {agent_server_image} does not exist in container registry, "
                     "make sure to build, push it, and make it public accessible before using remote workspace."
diff --git a/benchmarks/utils/build_utils.py b/benchmarks/utils/build_utils.py
index df5c10fa7..f61110729 100644
--- a/benchmarks/utils/build_utils.py
+++ b/benchmarks/utils/build_utils.py
@@ -27,7 +27,7 @@
     maybe_reset_buildkit,
 )
 from benchmarks.utils.constants import EVAL_AGENT_SERVER_IMAGE
-from benchmarks.utils.image_utils import image_exists, local_image_exists
+from benchmarks.utils.image_utils import local_image_exists, remote_image_exists
 from openhands.agent_server.docker.build import BuildOptions, TargetType, build
 from openhands.sdk import get_logger
 
@@ -300,7 +300,7 @@ def build_image(
     )
     for t in opts.all_tags:
         # Check if image exists or not
-        if image_exists(t):
+        if remote_image_exists(t):
             logger.info("Image %s already exists. Skipping build.", t)
             return BuildOutput(base_image=base_image, tags=[t], error=None)
     tags = build(opts)
diff --git a/benchmarks/utils/image_utils.py b/benchmarks/utils/image_utils.py
index bbbecb74f..467074cb9 100644
--- a/benchmarks/utils/image_utils.py
+++ b/benchmarks/utils/image_utils.py
@@ -120,12 +120,13 @@ def create_docker_workspace(
         )
 
 
-def image_exists(
+def remote_image_exists(
     image_ref: str,
     gh_username: str | None = None,
     gh_pat: str | None = None,  # GitHub PAT with read:packages for private GHCR
     docker_token: str | None = None,  # Docker Hub JWT if you already have one
 ) -> bool:
+    """Check if a Docker image exists in a remote registry."""
     registry, repo, ref = _parse(image_ref)
     headers = {"Accept": ACCEPT}
 
@@ -167,5 +168,5 @@ def image_exists(
     gh_user = sys.argv[2] if len(sys.argv) > 2 else None
     gh_pat = sys.argv[3] if len(sys.argv) > 3 else None
 
-    ok = image_exists(image, gh_username=gh_user, gh_pat=gh_pat)
+    ok = remote_image_exists(image, gh_username=gh_user, gh_pat=gh_pat)
     print(f"{image} -> {'✅ exists' if ok else '❌ not found or unauthorized'}")