diff --git a/tests/config/test_speculative_draft_hf_overrides.py b/tests/config/test_speculative_draft_hf_overrides.py
index e7014faa47d0..39ebd7924bf2 100644
--- a/tests/config/test_speculative_draft_hf_overrides.py
+++ b/tests/config/test_speculative_draft_hf_overrides.py
@@ -15,7 +15,7 @@
 from unittest.mock import MagicMock, patch
 
 import pytest
-from transformers import PreTrainedConfig
+from transformers import Glm5NextConfig, PreTrainedConfig
 
 from vllm.config.parallel import ParallelConfig
 from vllm.config.speculative import SpeculativeConfig
@@ -270,3 +270,21 @@ def test_mtp_index_share_override(
         speculative_config.draft_model_config.hf_config.index_share_for_mtp_iteration
         is expected
     )
+
+
+@pytest.mark.cpu_test
+def test_glm5_next_mtp_override_promotes_index_share():
+    """The flag lives on text_config, but the MTP proposers read the draft config."""
+    config = Glm5NextConfig(
+        architectures=["Glm5NextForConditionalGeneration"],
+        text_config={
+            "num_nextn_predict_layers": 1,
+            "index_share_for_mtp_iteration": True,
+        },
+    )
+
+    draft_config = SpeculativeConfig.hf_config_override(config)
+
+    assert draft_config.architectures == ["Glm5NextMTPModel"]
+    assert draft_config.n_predict == 1
+    assert draft_config.index_share_for_mtp_iteration is True
diff --git a/vllm/config/speculative.py b/vllm/config/speculative.py
index 4c61ead82dab..b916d7da2bed 100644
--- a/vllm/config/speculative.py
+++ b/vllm/config/speculative.py
@@ -1058,9 +1058,16 @@ def hf_config_override(hf_config: PreTrainedConfig) -> PreTrainedConfig:
             )
         if hf_config.model_type == "glm5_next":
             hf_config.model_type = "glm5_next_mtp"
-            n_predict = hf_config.get_text_config().num_nextn_predict_layers
+            text_config = hf_config.get_text_config()
+            share_mtp_indices = getattr(
+                text_config, "index_share_for_mtp_iteration", False
+            )
             hf_config.update(
-                {"n_predict": n_predict, "architectures": ["Glm5NextMTPModel"]}
+                {
+                    "n_predict": text_config.num_nextn_predict_layers,
+                    "architectures": ["Glm5NextMTPModel"],
+                    "index_share_for_mtp_iteration": share_mtp_indices,
+                }
             )
 
         return hf_config