fix(gemini): sweep hardcoded Gemini defaults to gemini-3.6-flash (fixes #32360)

agent/auxiliary_client.py

@@ -536,7 +536,7 @@ def _get_aux_model_for_provider(provider_id: str) -> str:
 # plus providers we intentionally keep pinned here (e.g. Anthropic predates
 # profiles). New providers should set default_aux_model on their profile instead.
 _API_KEY_PROVIDER_AUX_MODELS_FALLBACK: Dict[str, str] = {
-    "gemini": "gemini-3-flash-preview",
+    "gemini": "gemini-3.6-flash",
     "zai": "glm-4.5-flash",
     "kimi-coding": "kimi-k2-turbo-preview",
     "stepfun": "step-3.5-flash",
@@ -545,7 +545,7 @@ def _get_aux_model_for_provider(provider_id: str) -> str:
     "anthropic": "claude-haiku-4-5-20251001",
     "opencode-zen": "gemini-3-flash",
     "opencode-go": "glm-5",
-    "kilocode": "google/gemini-3-flash-preview",
+    "kilocode": "google/gemini-3.6-flash",
     "ollama-cloud": "nemotron-3-nano:30b",
     "tencent-tokenhub": "hy3-preview",
     # NB: no "deepinfra" entry — its aux model lives on the ProviderProfile
@@ -760,8 +760,8 @@ def _nous_extra_body() -> dict:
 auxiliary_is_nous: bool = False
 
 # Default auxiliary models per provider
-_OPENROUTER_MODEL = "google/gemini-3-flash-preview"
-_NOUS_MODEL = "google/gemini-3-flash-preview"
+_OPENROUTER_MODEL = "google/gemini-3.6-flash"
+_NOUS_MODEL = "google/gemini-3.6-flash"
 _NOUS_DEFAULT_BASE_URL = "https://inference-api.nousresearch.com/v1"
 _ANTHROPIC_DEFAULT_BASE_URL = "https://api.anthropic.com"
 _AUTH_JSON_PATH = get_hermes_home() / "auth.json"

agent/gemini_native_adapter.py

@@ -73,7 +73,7 @@ def probe_gemini_tier(
     api_key: str,
     base_url: str = DEFAULT_GEMINI_BASE_URL,
     *,
-    model: str = "gemini-2.5-flash",
+    model: str = "gemini-3.6-flash",
     timeout: float = 10.0,
 ) -> str:
     """Probe a Google AI Studio API key and return its tier.
@@ -154,8 +154,8 @@ def is_free_tier_quota_error(error_message: str) -> bool:
 
 
 _FREE_TIER_GUIDANCE = (
-    "\n\nYour Google API key is on the free tier (<= 250 requests/day for "
-    "gemini-2.5-flash). Hermes typically makes 3-10 API calls per user turn, "
+    "\n\nYour Google API key is on the free tier (a few hundred requests/day "
+    "for Gemini Flash models). Hermes typically makes 3-10 API calls per user turn, "
     "so the free tier is exhausted in a handful of messages and cannot sustain "
     "an agent session. Enable billing on your Google Cloud project and "
     "regenerate the key in a billing-enabled project: "
@@ -976,7 +976,7 @@ def _advance_stream_iterator(iterator: Iterator[_GeminiStreamChunk]) -> tuple[bo
     def _create_chat_completion(
         self,
         *,
-        model: str = "gemini-2.5-flash",
+        model: str = "gemini-3.6-flash",
         messages: Optional[List[Dict[str, Any]]] = None,
         stream: bool = False,
         tools: Any = None,

hermes_cli/setup.py

@@ -92,7 +92,7 @@ def _supports_same_provider_pool_setup(provider: str) -> bool:
     ],
     "gemini": [
         "gemini-3.1-pro-preview", "gemini-3-pro-preview",
-        "gemini-3-flash-preview", "gemini-3.1-flash-lite-preview",
+        "gemini-3.6-flash", "gemini-3.1-flash-lite-preview",
     ],
     "vertex": [
         "google/gemini-3.1-pro-preview", "google/gemini-3-pro-preview",

plugins/memory/hindsight/__init__.py

@@ -66,7 +66,7 @@
 _PROVIDER_DEFAULT_MODELS = {
     "openai": "gpt-4o-mini",
     "anthropic": "claude-haiku-4-5",
-    "gemini": "gemini-2.5-flash",
+    "gemini": "gemini-3.6-flash",
     "groq": "openai/gpt-oss-120b",
     "openrouter": "qwen/qwen3.5-9b",
     "minimax": "MiniMax-M2.7",

plugins/model-providers/kilocode/__init__.py

@@ -8,7 +8,7 @@
     aliases=("kilo-code", "kilo", "kilo-gateway"),
     env_vars=("KILOCODE_API_KEY",),
     base_url="https://api.kilo.ai/api/gateway",
-    default_aux_model="google/gemini-3-flash-preview",
+    default_aux_model="google/gemini-3.6-flash",
 )
 
 register_provider(kilocode)

plugins/model-providers/openrouter/__init__.py

@@ -204,7 +204,7 @@ def build_api_kwargs_extras(
         "anthropic/claude-sonnet-4.6",
         "openai/gpt-5.4",
         "deepseek/deepseek-chat",
-        "google/gemini-3-flash-preview",
+        "google/gemini-3.6-flash",
         "qwen/qwen3-plus",
     ),
 )

plugins/model-providers/vertex/__init__.py

@@ -69,7 +69,7 @@ def fetch_models(
     env_vars=(),  # OAuth2 via service account / ADC — not a static key env var
     base_url="https://aiplatform.googleapis.com",  # real base_url computed at runtime
     auth_type="vertex",
-    default_aux_model="google/gemini-3-flash-preview",
+    default_aux_model="google/gemini-3.6-flash",
 )
 
 register_provider(vertex)

tests/agent/test_auxiliary_client.py

@@ -10,6 +10,7 @@
 import pytest
 
 from agent.auxiliary_client import (
+    _NOUS_MODEL,
     get_text_auxiliary_client,
     get_available_vision_backends,
     resolve_vision_provider_client,
@@ -1804,7 +1805,7 @@ def select(self):
             client, model = _try_nous()
 
         assert client is not None
-        assert model == "google/gemini-3-flash-preview"
+        assert model == _NOUS_MODEL
         assert mock_openai.call_args.kwargs["api_key"] == pooled_token
         assert mock_openai.call_args.kwargs["base_url"] == "https://inference.pool.example/v1"
 
@@ -1851,7 +1852,7 @@ def try_refresh_current(self):
 
         assert pool.refreshed is True
         assert client is not None
-        assert model == "google/gemini-3-flash-preview"
+        assert model == _NOUS_MODEL
         assert mock_openai.call_args.kwargs["api_key"] == fresh_token
         assert mock_openai.call_args.kwargs["base_url"] == "https://inference.pool.example/v1"
 
@@ -1938,7 +1939,7 @@ def test_try_nous_falls_back_when_recommendation_lookup_raises(self):
             client, model = _try_nous()
 
         assert client is not None
-        assert model == "google/gemini-3-flash-preview"
+        assert model == _NOUS_MODEL
 
     def test_call_llm_retries_nous_after_401(self):
         class _Auth401(Exception):
@@ -2360,7 +2361,7 @@ def test_falls_back_to_default_when_portal_matches_stale(self, monkeypatch):
         )
         out = _refresh_nous_recommended_model(
             vision=True, stale_model="openai/gpt-5.4-mini")
-        assert out == "google/gemini-3-flash-preview"
+        assert out == _NOUS_MODEL
 
     def test_falls_back_to_default_when_portal_unavailable(self, monkeypatch):
         def _boom(**kw):
@@ -2369,17 +2370,17 @@ def _boom(**kw):
             "hermes_cli.models.get_nous_recommended_aux_model", _boom)
         out = _refresh_nous_recommended_model(
             vision=False, stale_model="some/dead-model")
-        assert out == "google/gemini-3-flash-preview"
+        assert out == _NOUS_MODEL
 
     def test_returns_none_when_no_distinct_alternative(self, monkeypatch):
         """When the failed model IS the default and the Portal has nothing
         else, there's no usable alternative."""
         monkeypatch.setattr(
             "hermes_cli.models.get_nous_recommended_aux_model",
-            lambda **kw: "google/gemini-3-flash-preview",
+            lambda **kw: _NOUS_MODEL,
         )
         out = _refresh_nous_recommended_model(
-            vision=False, stale_model="google/gemini-3-flash-preview")
+            vision=False, stale_model=_NOUS_MODEL)
         assert out is None
 
 

tests/run_agent/test_provider_parity.py

@@ -1022,15 +1022,15 @@ class TestAuxiliaryClientProviderPriority:
 
     def test_openrouter_always_wins(self, monkeypatch):
         monkeypatch.setenv("OPENROUTER_API_KEY", "or-key")
-        from agent.auxiliary_client import get_text_auxiliary_client
+        from agent.auxiliary_client import _OPENROUTER_MODEL, get_text_auxiliary_client
         with patch("agent.auxiliary_client.OpenAI") as mock:
             client, model = get_text_auxiliary_client()
-        assert model == "google/gemini-3-flash-preview"
+        assert model == _OPENROUTER_MODEL
         assert "openrouter" in str(mock.call_args.kwargs["base_url"]).lower()
 
     def test_nous_when_no_openrouter(self, monkeypatch):
         monkeypatch.delenv("OPENROUTER_API_KEY", raising=False)
-        from agent.auxiliary_client import get_text_auxiliary_client
+        from agent.auxiliary_client import _NOUS_MODEL, get_text_auxiliary_client
         nous_auth = {
             "access_token": _fake_invoke_jwt(),
             "scope": "inference:invoke",
@@ -1039,7 +1039,7 @@ def test_nous_when_no_openrouter(self, monkeypatch):
              patch("agent.auxiliary_client.OpenAI") as mock, \
              patch("hermes_cli.models.get_nous_recommended_aux_model", return_value=None):
             client, model = get_text_auxiliary_client()
-        assert model == "google/gemini-3-flash-preview"
+        assert model == _NOUS_MODEL
 
     def test_custom_endpoint_when_no_nous(self, monkeypatch):
         """Custom endpoint is used when no OpenRouter/Nous keys are available.