diff --git a/.github/workflows/opensearch.yml b/.github/workflows/opensearch.yml
index b91a419ffa..d4957d22f5 100644
--- a/.github/workflows/opensearch.yml
+++ b/.github/workflows/opensearch.yml
@@ -55,8 +55,8 @@ jobs:
         if: matrix.python-version == '3.10' && runner.os == 'Linux'
         run: hatch run docs
 
-      - name: Run tests
-        run: hatch run test:cov-retry
+      - name: Run tests (in parallel, using 4 cores)
+        run: hatch run test:cov-retry -n 4  # GA runner has 4 cores
 
       - name: Run unit tests with lowest direct dependencies
         run: |
diff --git a/integrations/opensearch/pyproject.toml b/integrations/opensearch/pyproject.toml
index 466863c445..e9652d1e60 100644
--- a/integrations/opensearch/pyproject.toml
+++ b/integrations/opensearch/pyproject.toml
@@ -58,6 +58,7 @@ dependencies = [
     "pytest-asyncio",
     "pytest-cov",
     "pytest-rerunfailures",
+    "pytest-xdist",
     "mypy",
     "pip",
     "boto3",
@@ -164,4 +165,3 @@ exclude_lines = ["no cov", "if __name__ == .__main__.:", "if TYPE_CHECKING:"]
 minversion = "6.0"
 markers = ["unit: unit tests", "integration: integration tests"]
 asyncio_mode = "auto"
-
diff --git a/integrations/opensearch/tests/conftest.py b/integrations/opensearch/tests/conftest.py
index 8d7e30bbc1..2ae8462faa 100644
--- a/integrations/opensearch/tests/conftest.py
+++ b/integrations/opensearch/tests/conftest.py
@@ -1,4 +1,5 @@
 import asyncio
+import uuid
 
 import pytest
 from haystack import Document
@@ -6,15 +7,23 @@
 from haystack_integrations.document_stores.opensearch.document_store import OpenSearchDocumentStore
 
 
+def _get_unique_index_name() -> str:
+    """
+    Generate a unique, valid OpenSearch index name for test isolation.
+
+    Each test gets its own index to enable parallel test execution without conflicts.
+    """
+    return f"test_{uuid.uuid4().hex}"
+
+
 @pytest.fixture
-def document_store(request):
+def document_store():
     """
     We use this document store for basic tests and for testing filters.
     `return_embedding` is set to True because in filters tests we compare embeddings.
     """
     hosts = ["https://localhost:9200"]
-    # Use a different index for each test so we can run them in parallel
-    index = f"{request.node.name}"
+    index = _get_unique_index_name()
 
     store = OpenSearchDocumentStore(
         hosts=hosts,
@@ -36,9 +45,9 @@ def document_store(request):
 
 
 @pytest.fixture
-def document_store_2(request):
+def document_store_2():
     hosts = ["https://localhost:9200"]
-    index = f"test_index_2_{request.node.name}"
+    index = f"test_index_2_{_get_unique_index_name()}"
 
     store = OpenSearchDocumentStore(
         hosts=hosts,
@@ -61,13 +70,13 @@ def document_store_2(request):
 
 
 @pytest.fixture
-def document_store_readonly(request):
+def document_store_readonly():
     """
     A document store that does not automatically create the underlying index.
     """
     hosts = ["https://localhost:9200"]
     # Use a different index for each test so we can run them in parallel
-    index = f"{request.node.name}"
+    index = _get_unique_index_name()
 
     store = OpenSearchDocumentStore(
         hosts=hosts,
@@ -90,13 +99,12 @@ def document_store_readonly(request):
 
 
 @pytest.fixture
-def document_store_embedding_dim_4_no_emb_returned(request):
+def document_store_embedding_dim_4_no_emb_returned():
     """
     A document store with embedding dimension 4 that does not return embeddings.
     """
     hosts = ["https://localhost:9200"]
-    # Use a different index for each test so we can run them in parallel
-    index = f"{request.node.name}"
+    index = _get_unique_index_name()
 
     store = OpenSearchDocumentStore(
         hosts=hosts,
@@ -113,15 +121,14 @@ def document_store_embedding_dim_4_no_emb_returned(request):
 
 
 @pytest.fixture
-def document_store_embedding_dim_4_no_emb_returned_faiss(request):
+def document_store_embedding_dim_4_no_emb_returned_faiss():
     """
     A document store with embedding dimension 4 that uses a FAISS engine with HNSW algorithm for vector search.
     We use this document store for testing efficient k-NN filtering according to
     https://opensearch.org/docs/latest/vector-search/filter-search-knn/efficient-knn-filtering/.
     """
     hosts = ["https://localhost:9200"]
-    # Use a different index for each test so we can run them in parallel
-    index = f"{request.node.name}"
+    index = _get_unique_index_name()
 
     store = OpenSearchDocumentStore(
         hosts=hosts,
diff --git a/integrations/opensearch/tests/test_document_store.py b/integrations/opensearch/tests/test_document_store.py
index 2f5976bb98..c75a5ca0a9 100644
--- a/integrations/opensearch/tests/test_document_store.py
+++ b/integrations/opensearch/tests/test_document_store.py
@@ -443,28 +443,6 @@ def test_write_documents_max_chunk_bytes(self, mock_bulk, document_store):
 
         assert mock_bulk.call_args.kwargs["max_chunk_bytes"] == DEFAULT_MAX_CHUNK_BYTES
 
-    @pytest.fixture
-    def document_store_embedding_dim_4_no_emb_returned(self, request):
-        """
-        This is the most basic requirement for the child class: provide
-        an instance of this document store so the base class can use it.
-        """
-        hosts = ["https://localhost:9200"]
-        # Use a different index for each test so we can run them in parallel
-        index = f"{request.node.name}"
-
-        store = OpenSearchDocumentStore(
-            hosts=hosts,
-            index=index,
-            http_auth=("admin", "admin"),
-            verify_certs=False,
-            embedding_dim=4,
-            return_embedding=False,
-            method={"space_type": "cosinesimil", "engine": "nmslib", "name": "hnsw"},
-        )
-        yield store
-        store._client.indices.delete(index=index, params={"ignore": [400, 404]})
-
     def test_embedding_retrieval_but_dont_return_embeddings_for_embedding_retrieval(
         self, document_store_embedding_dim_4_no_emb_returned: OpenSearchDocumentStore
     ):