diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py
index 7ea6fdee6a5..6e6f711f468 100644
--- a/litellm/proxy/proxy_server.py
+++ b/litellm/proxy/proxy_server.py
@@ -12768,6 +12768,7 @@ async def _fetch_db_models_for_search(
     size: int,
     sort_by: str | None,
     is_byok_outside_caller_teams: Callable[[dict[str, JsonValue]], bool],
+    model_name: str | None = None,
 ) -> tuple[list[dict[str, Any]], int]:
     """
     Run the bounded DB query that backs `/v2/model/info?search=`. Returns
@@ -12784,7 +12785,11 @@ async def _fetch_db_models_for_search(
     filter for `team_public_model_name` instead and keep the DB cost
     bounded by `search`.
     """
-    db_where_condition: Final[dict[str, Any]] = {"model_name": {"contains": search_lower, "mode": "insensitive"}}
+    exact_name_filter: Final[dict[str, Any]] = {} if model_name is None else {"AND": [{"model_name": model_name}]}
+    db_where_condition: Final[dict[str, Any]] = {
+        "model_name": {"contains": search_lower, "mode": "insensitive"},
+        **exact_name_filter,
+    }
     if db_model_ids_in_router:
         db_where_condition["model_id"] = {"not": {"in": list(db_model_ids_in_router)}}
 
@@ -12831,6 +12836,7 @@ async def _apply_search_filter_to_models(
     page: int = 1,
     size: int = 50,
     sort_by: str | None = None,
+    model_name: str | None = None,
 ) -> tuple[list[dict[str, Any]], int | None]:
     """
     Apply search filter to models, querying database for additional matching models.
@@ -12851,6 +12857,10 @@ async def _apply_search_filter_to_models(
         sort_by: Sort field. When set, results must be sorted across the
             full match set, so the DB fetch is capped at
             ``_SORTED_SEARCH_DB_FETCH_CAP`` instead of one page.
+        model_name: Exact ``model_name`` the caller already narrowed
+            ``all_models`` to (``?model=``). The DB query honours it too,
+            otherwise rows from other model groups leak into the result
+            and the count.
 
     Returns:
         Tuple of (filtered_models, total_count). total_count is None if not searching.
@@ -12920,6 +12930,7 @@ async def _apply_search_filter_to_models(
                 size=size,
                 sort_by=sort_by,
                 is_byok_outside_caller_teams=_is_byok_outside_caller_teams,
+                model_name=model_name,
             )
             search_total_count = router_models_count + db_models_total_count
         except Exception as e:
@@ -13485,6 +13496,7 @@ async def model_info_v2(
             page=page,
             size=size,
             sort_by=sortBy,
+            model_name=model,
         )
 
     if user_models_only:
diff --git a/tests/test_litellm/proxy/test_proxy_server.py b/tests/test_litellm/proxy/test_proxy_server.py
index 8e470cc663b..fa8d6d8df41 100644
--- a/tests/test_litellm/proxy/test_proxy_server.py
+++ b/tests/test_litellm/proxy/test_proxy_server.py
@@ -2126,6 +2126,43 @@ async def test_apply_search_filter_bounds_db_fetch_by_page_and_cap():
     assert take < 10_000, "sorted search must cap below the full match set"
 
 
+@pytest.mark.asyncio
+async def test_apply_search_filter_honours_exact_model_name_in_db_query():
+    """
+    `/v2/model/info?model=<group>&search=<term>`: the router list is already
+    narrowed to the exact group, so the DB count and fetch must be too, or
+    other groups' rows leak into the page and inflate total_count.
+    """
+    from litellm.proxy.proxy_server import _apply_search_filter_to_models
+
+    prisma_client = MagicMock()
+    prisma_client.db.litellm_proxymodeltable.count = AsyncMock(return_value=0)
+    prisma_client.db.litellm_proxymodeltable.find_many = AsyncMock(return_value=[])
+    proxy_config = MagicMock()
+    proxy_config.decrypt_model_list_from_db = lambda rows: []
+
+    await _apply_search_filter_to_models(
+        all_models=[],
+        search="sonnet",
+        prisma_client=prisma_client,
+        proxy_config=proxy_config,
+        model_name="anthropic-sonnet-5",
+    )
+    where = prisma_client.db.litellm_proxymodeltable.count.call_args.kwargs["where"]
+    assert where["AND"] == [{"model_name": "anthropic-sonnet-5"}]
+    assert where["model_name"] == {"contains": "sonnet", "mode": "insensitive"}
+    assert prisma_client.db.litellm_proxymodeltable.find_many.call_args.kwargs["where"] == where
+
+    prisma_client.db.litellm_proxymodeltable.count.reset_mock()
+    await _apply_search_filter_to_models(
+        all_models=[],
+        search="sonnet",
+        prisma_client=prisma_client,
+        proxy_config=proxy_config,
+    )
+    assert "AND" not in prisma_client.db.litellm_proxymodeltable.count.call_args.kwargs["where"]
+
+
 @pytest.mark.asyncio
 async def test_filter_models_by_team_id_excludes_viewer_direct_access():
     """