diff --git a/agent/models_dev.py b/agent/models_dev.py
index 42c8925ffe..cc4dbf0be4 100644
--- a/agent/models_dev.py
+++ b/agent/models_dev.py
@@ -420,7 +420,10 @@ def list_provider_models(provider: str) -> List[str]:
     models = _get_provider_models(provider)
     if models is None:
         return []
-    return list(models.keys())
+    return [
+        mid for mid in models.keys()
+        if not _should_hide_from_provider_catalog(provider, mid)
+    ]
 
 
 # Patterns that indicate non-agentic or noise models (TTS, embedding,
@@ -432,6 +435,29 @@ _NOISE_PATTERNS: re.Pattern = re.compile(
     re.IGNORECASE,
 )
 
+# Google-hosted Gemma models currently have very low TPM quotas for agent-style
+# traffic (for example 15K/16K TPM tiers in AI Studio) and are not practical as
+# normal Hermes picks even though they advertise large context windows. Keep the
+# capability metadata available for direct/manual use, but hide them from the
+# Gemini model catalogs we surface in setup and model selection.
+_GOOGLE_GEMMA_HIDDEN_MODELS = frozenset({
+    "gemma-4-31b-it",
+    "gemma-4-26b-a4b-it",
+    "gemma-3-1b",
+    "gemma-3-2b",
+    "gemma-3-4b",
+    "gemma-3-12b",
+    "gemma-3-27b",
+})
+
+
+def _should_hide_from_provider_catalog(provider: str, model_id: str) -> bool:
+    provider_lower = (provider or "").strip().lower()
+    model_lower = (model_id or "").strip().lower()
+    if provider_lower in {"gemini", "google"} and model_lower in _GOOGLE_GEMMA_HIDDEN_MODELS:
+        return True
+    return False
+
 
 def list_agentic_models(provider: str) -> List[str]:
     """Return model IDs suitable for agentic use from models.dev.
@@ -448,6 +474,8 @@ def list_agentic_models(provider: str) -> List[str]:
     for mid, entry in models.items():
         if not isinstance(entry, dict):
             continue
+        if _should_hide_from_provider_catalog(provider, mid):
+            continue
         if not entry.get("tool_call", False):
             continue
         if _NOISE_PATTERNS.search(mid):
@@ -582,5 +610,3 @@ def get_model_info(
             return _parse_model_info(mid, mdata, mdev_id)
 
     return None
-
-
diff --git a/hermes_cli/models.py b/hermes_cli/models.py
index cbbeef62d4..a0d7c2220c 100644
--- a/hermes_cli/models.py
+++ b/hermes_cli/models.py
@@ -133,8 +133,6 @@ _PROVIDER_MODELS: dict[str, list[str]] = {
         "gemini-2.5-pro",
         "gemini-2.5-flash",
         "gemini-2.5-flash-lite",
-        # Gemma open models (also served via AI Studio)
-        "gemma-4-31b-it",
     ],
     "google-gemini-cli": [
         "gemini-2.5-pro",
diff --git a/hermes_cli/setup.py b/hermes_cli/setup.py
index 8770386b73..8f6b633c6a 100644
--- a/hermes_cli/setup.py
+++ b/hermes_cli/setup.py
@@ -91,7 +91,6 @@ _DEFAULT_PROVIDER_MODELS = {
     "gemini": [
         "gemini-3.1-pro-preview", "gemini-3-flash-preview", "gemini-3.1-flash-lite-preview",
         "gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite",
-        "gemma-4-31b-it",
     ],
     "zai": ["glm-5.1", "glm-5", "glm-4.7", "glm-4.5", "glm-4.5-flash"],
     "kimi-coding": ["kimi-k2.5", "kimi-k2-thinking", "kimi-k2-turbo-preview"],
diff --git a/tests/hermes_cli/test_gemini_provider.py b/tests/hermes_cli/test_gemini_provider.py
index fd16e825d1..7632f7691c 100644
--- a/tests/hermes_cli/test_gemini_provider.py
+++ b/tests/hermes_cli/test_gemini_provider.py
@@ -130,7 +130,7 @@ class TestGeminiModelCatalog:
         models = _PROVIDER_MODELS["gemini"]
         assert "gemini-2.5-pro" in models
         assert "gemini-2.5-flash" in models
-        assert "gemma-4-31b-it" in models
+        assert "gemma-4-31b-it" not in models
 
     def test_provider_models_has_3x(self):
         models = _PROVIDER_MODELS["gemini"]
@@ -313,9 +313,28 @@ class TestGeminiModelsDev:
             result = list_agentic_models("gemini")
         assert "gemini-3-flash-preview" in result
         assert "gemini-2.5-pro" in result
-        assert "gemma-4-31b-it" in result
+        assert "gemma-4-31b-it" not in result
         # Filtered out:
         assert "gemini-embedding-001" not in result      # no tool_call
         assert "gemini-2.5-flash-preview-tts" not in result  # no tool_call
         assert "gemini-live-2.5-flash" not in result     # noise: live-
         assert "gemini-2.5-flash-preview-04-17" not in result  # noise: dated preview
+
+    def test_list_provider_models_hides_low_tpm_google_gemmas(self):
+        mock_data = {
+            "google": {
+                "models": {
+                    "gemini-2.5-pro": {},
+                    "gemma-4-31b-it": {},
+                    "gemma-3-1b": {},
+                }
+            }
+        }
+        with patch("agent.models_dev.fetch_models_dev", return_value=mock_data):
+            from agent.models_dev import list_provider_models
+
+            result = list_provider_models("gemini")
+
+        assert "gemini-2.5-pro" in result
+        assert "gemma-4-31b-it" not in result
+        assert "gemma-3-1b" not in result