diff --git a/backend/.env.example b/backend/.env.example
index b772481cb..c2efebc17 100644
--- a/backend/.env.example
+++ b/backend/.env.example
@@ -27,7 +27,7 @@ GOOGLE_APPLICATION_CREDENTIALS=keys/credentials.json
 # (e.g. us-central1) — these models are not published in the global publisher catalog.
 VERTEX_API_EMBEDDINGS_REGION=us-central1
 # Region for generative-AI calls (Gemini etc.). Can be a regional location or "global".
-VERTEX_API_GEN_AI_REGION=us-central1
+VERTEX_API_GEN_AI_REGION=global
 EMBEDDINGS_SERVICE_NAME=GOOGLE-VERTEX-AI
 # See https://cloud.google.com/vertex-ai/generative-ai/docs/embeddings/get-text-embeddings#supported-models e.g.:
 #   text-embedding-005
@@ -104,11 +104,13 @@ LLM_PROVIDER=gemini
 #   LLM_REASONING_MODEL       — classifiers, routing and extraction that need good reasoning
 #   LLM_DEEP_REASONING_MODEL  — the strongest reasoning, e.g. evaluation judges
 # When a tier is unset:
-#   gemini:    uses its own default (gemini-2.5-flash-lite / gemini-2.5-flash / gemini-2.5-pro)
+#   gemini:    uses its own default (gemini-3.5-flash-lite / gemini-3.8-flash / gemini-3.1-pro-preview)
 #   anthropic: falls back to the next tier down that is set
 #              (LLM_DEEP_REASONING_MODEL -> LLM_REASONING_MODEL -> LLM_DEFAULT_MODEL), then to claude-sonnet-4-6
 #   ollama:    same as anthropic, then to qwen2.5:7b
 # To use one model for everything with a non-Gemini provider, set only LLM_DEFAULT_MODEL.
+# Gemini models are called through the google-genai Interactions API, which only supports
+# gemini-3.5-flash-lite, gemini-3.8-flash and gemini-3.1-pro-preview (the 2.5 models are rejected).
 # LLM_DEFAULT_MODEL=
 # LLM_REASONING_MODEL=
 # LLM_DEEP_REASONING_MODEL=
diff --git a/backend/README.md b/backend/README.md
index df0cc6dee..0604a9444 100644
--- a/backend/README.md
+++ b/backend/README.md
@@ -567,13 +567,14 @@ For more details, see [app/agent/preference_elicitation_agent/README.md](app/age
 ### AI & Embeddings
 
 - `VERTEX_API_EMBEDDINGS_REGION` - Vertex AI region for embedding models (must be a regional location, e.g. `us-central1`)
-- `VERTEX_API_GEN_AI_REGION` - Vertex AI region for generative-AI calls (regional or `global`, default: `us-central1`)
+- `VERTEX_API_GEN_AI_REGION` - Vertex AI region for generative-AI calls (regional or `global`, default: `global`). The default and reasoning Gemini 3.x models are only available in `global`
 - `EMBEDDINGS_SERVICE_NAME` - Embeddings service provider
 - `EMBEDDINGS_MODEL_NAME` - Model for generating embeddings
 - `LLM_PROVIDER` - LLM backend for all agents: `gemini` (default), `anthropic` or `ollama`
-- `LLM_DEFAULT_MODEL` - Model for the default tier (gemini default: `gemini-2.5-flash-lite`)
-- `LLM_REASONING_MODEL` - Model for the reasoning tier (gemini default: `gemini-2.5-flash`)
-- `LLM_DEEP_REASONING_MODEL` - Model for the deep reasoning tier (gemini default: `gemini-2.5-pro`)
+- `LLM_DEFAULT_MODEL` - Model for the default tier (gemini default: `gemini-3.5-flash-lite`)
+- `LLM_REASONING_MODEL` - Model for the reasoning tier (gemini default: `gemini-3.8-flash`)
+- `LLM_DEEP_REASONING_MODEL` - Model for the deep reasoning tier (gemini default: `gemini-3.1-pro-preview`)
+  - Gemini models are called through the google-genai Interactions API, which only supports `gemini-3.5-flash-lite`, `gemini-3.8-flash` and `gemini-3.1-pro-preview` (the 2.5 models are rejected)
   - For `anthropic` and `ollama`, an unset tier falls back to the next tier down that is set (`LLM_DEEP_REASONING_MODEL` → `LLM_REASONING_MODEL` → `LLM_DEFAULT_MODEL`), then to the provider default (`claude-sonnet-4-6` / `qwen2.5:7b`)
 - `ANTHROPIC_API_KEY` - Anthropic API key, required when `LLM_PROVIDER=anthropic`
 - `OLLAMA_BASE_URL` - Ollama server URL when `LLM_PROVIDER=ollama` (default: `http://localhost:11434`)
diff --git a/backend/app/agent/config.py b/backend/app/agent/config.py
index 19141dc9e..f147b29de 100644
--- a/backend/app/agent/config.py
+++ b/backend/app/agent/config.py
@@ -2,9 +2,9 @@
 from typing import Literal, Final
 
 Model = Literal[
-    "gemini-2.5-flash-lite",
-    "gemini-2.5-flash",
-    "gemini-2.5-pro",
+    "gemini-3.5-flash-lite",
+    "gemini-3.8-flash",
+    "gemini-3.1-pro-preview",
 ]
 
 
@@ -39,10 +39,12 @@ class ModelTier(StrEnum):
 
 
 GEMINI_DEFAULT_MODELS: Final[dict[ModelTier, Model]] = {
-    ModelTier.DEFAULT: "gemini-2.5-flash-lite",
-    ModelTier.REASONING: "gemini-2.5-flash",
-    ModelTier.DEEP_REASONING: "gemini-2.5-pro",
+    ModelTier.DEFAULT: "gemini-3.5-flash-lite",
+    ModelTier.REASONING: "gemini-3.8-flash",
+    ModelTier.DEEP_REASONING: "gemini-3.1-pro-preview",
 }
 """
 The Gemini model used for each tier when the tier's environment variable is not set.
+The Gemini 3.x models are only served from the Vertex AI `global` location, see `VERTEX_API_GEN_AI_REGION`.
+These are models the google-genai Interactions API supports, it rejects the Gemini 2.5 models.
 """
diff --git a/backend/app/app_config.py b/backend/app/app_config.py
index ddf545fcb..a9d3c79e0 100644
--- a/backend/app/app_config.py
+++ b/backend/app/app_config.py
@@ -143,7 +143,7 @@ class ApplicationConfig(BaseModel):
     llm_default_model: Optional[str] = None
     """
     The model for the default tier (ModelTier.DEFAULT) of the selected provider.
-    When unset, each provider uses its own default (gemini: gemini-2.5-flash-lite,
+    When unset, each provider uses its own default (gemini: gemini-3.5-flash-lite,
     anthropic: claude-sonnet-4-6, ollama: qwen2.5:7b).
     Set via LLM_DEFAULT_MODEL environment variable.
     """
@@ -151,14 +151,14 @@ class ApplicationConfig(BaseModel):
     llm_reasoning_model: Optional[str] = None
     """
     The model for the reasoning tier (ModelTier.REASONING) of the selected provider.
-    When unset, gemini uses gemini-2.5-flash; other providers fall back to llm_default_model.
+    When unset, gemini uses gemini-3.8-flash; other providers fall back to llm_default_model.
     Set via LLM_REASONING_MODEL environment variable.
     """
 
     llm_deep_reasoning_model: Optional[str] = None
     """
     The model for the deep reasoning tier (ModelTier.DEEP_REASONING) of the selected provider.
-    When unset, gemini uses gemini-2.5-pro; other providers fall back to llm_reasoning_model, then llm_default_model.
+    When unset, gemini uses gemini-3.1-pro-preview; other providers fall back to llm_reasoning_model, then llm_default_model.
     Set via LLM_DEEP_REASONING_MODEL environment variable.
     """
 
diff --git a/backend/common_libs/llm/models_utils.py b/backend/common_libs/llm/models_utils.py
index 6fffa0185..de21182f1 100644
--- a/backend/common_libs/llm/models_utils.py
+++ b/backend/common_libs/llm/models_utils.py
@@ -24,8 +24,9 @@
 
 DEFAULT_VERTEX_API_GEN_AI_REGION = os.getenv("VERTEX_API_GEN_AI_REGION")
 if not DEFAULT_VERTEX_API_GEN_AI_REGION:
-    logging.warning("VERTEX_API_GEN_AI_REGION is not set. Using 'us-central1' as the default region.")
-    DEFAULT_VERTEX_API_GEN_AI_REGION = "us-central1"
+    # The Gemini 3.x models are only served from the global location, regional locations return a 404 for them.
+    logging.warning("VERTEX_API_GEN_AI_REGION is not set. Using 'global' as the default region.")
+    DEFAULT_VERTEX_API_GEN_AI_REGION = "global"
 else:
     logging.info("Default Vertex AI gen-AI region is %s", DEFAULT_VERTEX_API_GEN_AI_REGION)
 
@@ -167,7 +168,7 @@ def resolve_model_name(*, tier: ModelTier, provider: "LLMProvider") -> str:
             ModelTier.REASONING: app_config.llm_reasoning_model,
             ModelTier.DEEP_REASONING: app_config.llm_deep_reasoning_model,
         }
-        # Gemini falls back to its own 2.5 default for the tier rather than to another tier's model.
+        # Gemini falls back to its own default for the tier rather than to another tier's model.
         tiers_to_try = [tier] if provider == "gemini" else [tier, *_NON_GEMINI_FALLBACK_TIERS[tier]]
         for candidate_tier in tiers_to_try:
             if configured_models[candidate_tier]:
diff --git a/backend/common_libs/llm/test_model_resolution.py b/backend/common_libs/llm/test_model_resolution.py
index c4ebf7d64..ac741db3d 100644
--- a/backend/common_libs/llm/test_model_resolution.py
+++ b/backend/common_libs/llm/test_model_resolution.py
@@ -27,11 +27,11 @@ def _set_llm_config(config: ApplicationConfig, *,
 
 class TestResolveModelName:
     @pytest.mark.parametrize("given_tier, expected_model", [
-        (ModelTier.DEFAULT, "gemini-2.5-flash-lite"),
-        (ModelTier.REASONING, "gemini-2.5-flash"),
-        (ModelTier.DEEP_REASONING, "gemini-2.5-pro"),
+        (ModelTier.DEFAULT, "gemini-3.5-flash-lite"),
+        (ModelTier.REASONING, "gemini-3.8-flash"),
+        (ModelTier.DEEP_REASONING, "gemini-3.1-pro-preview"),
     ])
-    def test_gemini_uses_gemini_2_5_defaults_when_no_model_is_set(self, setup_application_config,
+    def test_gemini_uses_gemini_defaults_when_no_model_is_set(self, setup_application_config,
                                                                   given_tier: ModelTier, expected_model: str):
         # GIVEN the gemini provider is configured without any tier models
         _set_llm_config(setup_application_config, provider="gemini")
@@ -39,7 +39,7 @@ def test_gemini_uses_gemini_2_5_defaults_when_no_model_is_set(self, setup_applic
         # WHEN the model name for the given tier is resolved
         actual_model = resolve_model_name(tier=given_tier, provider="gemini")
 
-        # THEN expect the gemini 2.5 default for the tier
+        # THEN expect the gemini default for the tier
         assert actual_model == expected_model
 
     @pytest.mark.parametrize("given_tier, expected_model", [
@@ -63,9 +63,9 @@ def test_uses_the_tier_model_when_it_is_set(self, setup_application_config,
         assert actual_model == expected_model
 
     @pytest.mark.parametrize("given_reasoning_model, given_tier, expected_model", [
-        (None, ModelTier.REASONING, "gemini-2.5-flash"),
-        (None, ModelTier.DEEP_REASONING, "gemini-2.5-pro"),
-        ("given-reasoning", ModelTier.DEEP_REASONING, "gemini-2.5-pro"),
+        (None, ModelTier.REASONING, "gemini-3.8-flash"),
+        (None, ModelTier.DEEP_REASONING, "gemini-3.1-pro-preview"),
+        ("given-reasoning", ModelTier.DEEP_REASONING, "gemini-3.1-pro-preview"),
     ])
     def test_gemini_does_not_fall_back_to_another_tier(self, setup_application_config,
                                                        given_reasoning_model: Optional[str],
@@ -78,7 +78,7 @@ def test_gemini_does_not_fall_back_to_another_tier(self, setup_application_confi
         # WHEN the model name for a tier without a model is resolved
         actual_model = resolve_model_name(tier=given_tier, provider="gemini")
 
-        # THEN expect the gemini 2.5 default for the tier, not another tier's model
+        # THEN expect the gemini default for the tier, not another tier's model
         assert actual_model == expected_model
 
     @pytest.mark.parametrize("given_tier", list(ModelTier))
@@ -144,9 +144,9 @@ def test_non_gemini_default_tier_does_not_fall_back_to_a_higher_tier(self, setup
         assert actual_model == expected_model
 
     @pytest.mark.parametrize("given_tier, expected_model", [
-        (ModelTier.DEFAULT, "gemini-2.5-flash-lite"),
-        (ModelTier.REASONING, "gemini-2.5-flash"),
-        (ModelTier.DEEP_REASONING, "gemini-2.5-pro"),
+        (ModelTier.DEFAULT, "gemini-3.5-flash-lite"),
+        (ModelTier.REASONING, "gemini-3.8-flash"),
+        (ModelTier.DEEP_REASONING, "gemini-3.1-pro-preview"),
     ])
     def test_ignores_the_models_of_another_provider(self, setup_application_config,
                                                     given_tier: ModelTier, expected_model: str):
@@ -159,13 +159,13 @@ def test_ignores_the_models_of_another_provider(self, setup_application_config,
         # WHEN the model name for the given tier is resolved for gemini (e.g. a gemini-only call site)
         actual_model = resolve_model_name(tier=given_tier, provider="gemini")
 
-        # THEN expect the gemini 2.5 default for the tier
+        # THEN expect the gemini default for the tier
         assert actual_model == expected_model
 
     @pytest.mark.parametrize("given_tier, expected_model", [
-        (ModelTier.DEFAULT, "gemini-2.5-flash-lite"),
-        (ModelTier.REASONING, "gemini-2.5-flash"),
-        (ModelTier.DEEP_REASONING, "gemini-2.5-pro"),
+        (ModelTier.DEFAULT, "gemini-3.5-flash-lite"),
+        (ModelTier.REASONING, "gemini-3.8-flash"),
+        (ModelTier.DEEP_REASONING, "gemini-3.1-pro-preview"),
     ])
     def test_uses_the_defaults_when_the_application_config_is_not_set(self, given_tier: ModelTier,
                                                                       expected_model: str):
@@ -175,7 +175,7 @@ def test_uses_the_defaults_when_the_application_config_is_not_set(self, given_ti
         # WHEN the model name for the given tier is resolved
         actual_model = resolve_model_name(tier=given_tier, provider="gemini")
 
-        # THEN expect the gemini 2.5 default for the tier
+        # THEN expect the gemini default for the tier
         assert actual_model == expected_model
 
 
@@ -214,7 +214,7 @@ def test_explicit_model_name_is_used_as_is(self, setup_application_config):
 
 class TestGeminiGenerativeLLM:
     @pytest.mark.parametrize("given_config, expected_model", [
-        (LLMConfig(), "gemini-2.5-flash-lite"),
+        (LLMConfig(), "gemini-3.5-flash-lite"),
         (LLMConfig(model_tier=ModelTier.DEEP_REASONING), "given-deep-reasoning"),
         (LLMConfig(model_tier=ModelTier.DEEP_REASONING, language_model_name="given-explicit-model"),
          "given-explicit-model"),
diff --git a/backend/evaluation_tests/agent_director/llm_agent_director_scripted_user_test.py b/backend/evaluation_tests/agent_director/llm_agent_director_scripted_user_test.py
index 371862fa0..acf3547d8 100644
--- a/backend/evaluation_tests/agent_director/llm_agent_director_scripted_user_test.py
+++ b/backend/evaluation_tests/agent_director/llm_agent_director_scripted_user_test.py
@@ -104,7 +104,7 @@ async def agent_director_exec(caplog: LogCaptureFixture, test_case: ScriptedUser
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_user_says_all_the_time_yes(caplog: LogCaptureFixture,
                                           setup_agent_director: Awaitable[tuple[ConversationMemoryManager, Callable[
@@ -142,7 +142,7 @@ async def test_user_says_all_the_time_yes(caplog: LogCaptureFixture,
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_user_talks_about_occupations(caplog: LogCaptureFixture,
                                             setup_agent_director: Awaitable[tuple[ConversationMemoryManager, Callable[
@@ -206,7 +206,7 @@ class AgentState:
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_argentina_counseling_flow(caplog: LogCaptureFixture,
                                          setup_agent_director: Awaitable[tuple[ConversationMemoryManager, Callable[
diff --git a/backend/evaluation_tests/career_explorer_agent/career_explorer_agent_scripted_user_test.py b/backend/evaluation_tests/career_explorer_agent/career_explorer_agent_scripted_user_test.py
index 0c3937760..b0b489b26 100644
--- a/backend/evaluation_tests/career_explorer_agent/career_explorer_agent_scripted_user_test.py
+++ b/backend/evaluation_tests/career_explorer_agent/career_explorer_agent_scripted_user_test.py
@@ -183,7 +183,7 @@ def career_explorer_config_with_sectors(setup_multi_locale_app_config):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", TEST_CASES, ids=[tc.name for tc in TEST_CASES])
 async def test_career_explorer_sector_content(
     evals_setup, setup_multi_locale_app_config, career_explorer_config_with_sectors, test_case: SectorContentTestCase
@@ -239,7 +239,7 @@ async def test_career_explorer_sector_content(
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize(
     "test_case", TEST_CASES_GENERAL_KNOWLEDGE, ids=[tc.name for tc in TEST_CASES_GENERAL_KNOWLEDGE]
 )
diff --git a/backend/evaluation_tests/career_explorer_agent/non_priority_sector_explorer_eval_test.py b/backend/evaluation_tests/career_explorer_agent/non_priority_sector_explorer_eval_test.py
index 4b0779019..2cfeecb35 100644
--- a/backend/evaluation_tests/career_explorer_agent/non_priority_sector_explorer_eval_test.py
+++ b/backend/evaluation_tests/career_explorer_agent/non_priority_sector_explorer_eval_test.py
@@ -54,7 +54,7 @@ def career_explorer_config_with_sectors(setup_multi_locale_app_config):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", NON_PRIORITY_EXPLORER_TEST_CASES, ids=[tc.name for tc in NON_PRIORITY_EXPLORER_TEST_CASES])
 async def test_non_priority_sector_explorer_web_search(
     evals_setup,
diff --git a/backend/evaluation_tests/career_explorer_agent/priority_sector_explorer_eval_test.py b/backend/evaluation_tests/career_explorer_agent/priority_sector_explorer_eval_test.py
index 18186bff9..a3df1c11a 100644
--- a/backend/evaluation_tests/career_explorer_agent/priority_sector_explorer_eval_test.py
+++ b/backend/evaluation_tests/career_explorer_agent/priority_sector_explorer_eval_test.py
@@ -104,7 +104,7 @@ def career_explorer_config_with_sectors(setup_multi_locale_app_config):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", PRIORITY_EXPLORER_TEST_CASES, ids=[tc.name for tc in PRIORITY_EXPLORER_TEST_CASES])
 async def test_priority_sector_explorer_rag_content(
     evals_setup,
diff --git a/backend/evaluation_tests/career_explorer_agent/sector_relevance_classifier_eval_test.py b/backend/evaluation_tests/career_explorer_agent/sector_relevance_classifier_eval_test.py
index c4b0bc296..ec6a30312 100644
--- a/backend/evaluation_tests/career_explorer_agent/sector_relevance_classifier_eval_test.py
+++ b/backend/evaluation_tests/career_explorer_agent/sector_relevance_classifier_eval_test.py
@@ -70,7 +70,7 @@ def career_explorer_config_with_sectors(setup_multi_locale_app_config):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", CLASSIFIER_TEST_CASES, ids=[tc.name for tc in CLASSIFIER_TEST_CASES])
 async def test_sector_relevance_classifier(
     evals_setup,
@@ -116,7 +116,7 @@ async def test_sector_relevance_classifier(
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 async def test_classifier_extracts_multiple_sectors(
     evals_setup,
     career_explorer_config_with_sectors,
diff --git a/backend/evaluation_tests/career_readiness_agent/career_readiness_agent_scripted_user_test.py b/backend/evaluation_tests/career_readiness_agent/career_readiness_agent_scripted_user_test.py
index b2a274bd9..b575e7ab4 100644
--- a/backend/evaluation_tests/career_readiness_agent/career_readiness_agent_scripted_user_test.py
+++ b/backend/evaluation_tests/career_readiness_agent/career_readiness_agent_scripted_user_test.py
@@ -246,7 +246,7 @@ def _save_conversation_output(
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", TEST_CASES, ids=[tc.name for tc in TEST_CASES])
 async def test_career_readiness_scripted(evals_setup, setup_multi_locale_app_config,
                                          test_case: CareerReadinessTestCase):
@@ -285,7 +285,7 @@ async def test_career_readiness_scripted(evals_setup, setup_multi_locale_app_con
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", EDGE_CASE_TEST_CASES, ids=[tc.name for tc in EDGE_CASE_TEST_CASES])
 async def test_career_readiness_edge_cases(evals_setup, setup_multi_locale_app_config,
                                             test_case: CareerReadinessEdgeCaseTestCase):
diff --git a/backend/evaluation_tests/collect_experiences_agent/_data_extraction_llm_test.py b/backend/evaluation_tests/collect_experiences_agent/_data_extraction_llm_test.py
index f2c0e98aa..a3a276b8b 100644
--- a/backend/evaluation_tests/collect_experiences_agent/_data_extraction_llm_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/_data_extraction_llm_test.py
@@ -999,7 +999,7 @@ class _TestCaseDataExtraction(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases_data_extraction),
                          ids=[case.name for case in get_test_cases_to_run(test_cases_data_extraction)])
diff --git a/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_scripted_user_test.py b/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_scripted_user_test.py
index 781006732..05e5e327f 100644
--- a/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_scripted_user_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_scripted_user_test.py
@@ -67,7 +67,7 @@ async def collect_experiences_exec(caplog: LogCaptureFixture, test_case: Scripte
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_kenyan_user_multiple_experiences_transition_issues(caplog: LogCaptureFixture,
                                                                   setup_collect_experiences_agent: Awaitable[tuple[
@@ -201,7 +201,7 @@ async def test_stuck_after_recap_confirmation(caplog: LogCaptureFixture,
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_quick_reply_options_appear_at_yes_no_questions(caplog: LogCaptureFixture,
                                                               setup_collect_experiences_agent: Awaitable[tuple[
diff --git a/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_simulated_user_test.py b/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_simulated_user_test.py
index a2d7252da..36387738d 100644
--- a/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_simulated_user_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/collect_experiences_agent_simulated_user_test.py
@@ -21,7 +21,7 @@
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
diff --git a/backend/evaluation_tests/collect_experiences_agent/data_extraction/entity_extraction_tool_test.py b/backend/evaluation_tests/collect_experiences_agent/data_extraction/entity_extraction_tool_test.py
index 9926a12f3..a27eba243 100644
--- a/backend/evaluation_tests/collect_experiences_agent/data_extraction/entity_extraction_tool_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/data_extraction/entity_extraction_tool_test.py
@@ -385,7 +385,7 @@ class EntityExtractionToolTestCase(CompassTestCase):
 
 @pytest.mark.asyncio
 @pytest.mark.repeat(3)
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
 async def test_entity_extraction_tool(test_case: EntityExtractionToolTestCase, caplog):
diff --git a/backend/evaluation_tests/collect_experiences_agent/data_extraction/intent_analyzer_llm_test.py b/backend/evaluation_tests/collect_experiences_agent/data_extraction/intent_analyzer_llm_test.py
index 0600011b7..94cc6f179 100644
--- a/backend/evaluation_tests/collect_experiences_agent/data_extraction/intent_analyzer_llm_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/data_extraction/intent_analyzer_llm_test.py
@@ -264,7 +264,7 @@ class IntentAnalyzerToolTestCase(CompassTestCase):
 
 @pytest.mark.asyncio
 @pytest.mark.repeat(3)
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
 async def test_intent_analyzer_tool(test_case: IntentAnalyzerToolTestCase, caplog):
diff --git a/backend/evaluation_tests/collect_experiences_agent/data_extraction/temporal_classifier_test.py b/backend/evaluation_tests/collect_experiences_agent/data_extraction/temporal_classifier_test.py
index 6ab84054f..1dc11b54e 100644
--- a/backend/evaluation_tests/collect_experiences_agent/data_extraction/temporal_classifier_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/data_extraction/temporal_classifier_test.py
@@ -298,7 +298,7 @@ class TemporalAndWorkTypeClassifierToolTestCase(CompassTestCase):
 
 @pytest.mark.asyncio
 @pytest.mark.repeat(3)
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
 async def test_temporal_and_work_type_classification(test_case: TemporalAndWorkTypeClassifierToolTestCase, caplog,
diff --git a/backend/evaluation_tests/collect_experiences_agent/data_extraction/transition_decision_tool_test.py b/backend/evaluation_tests/collect_experiences_agent/data_extraction/transition_decision_tool_test.py
index 29a817a4e..aae06095a 100644
--- a/backend/evaluation_tests/collect_experiences_agent/data_extraction/transition_decision_tool_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/data_extraction/transition_decision_tool_test.py
@@ -448,7 +448,7 @@ class TransitionDecisionToolTestCase(CompassTestCase):
 
 @pytest.mark.asyncio
 @pytest.mark.repeat(3)
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
 async def test_transition_decision_tool(test_case: TransitionDecisionToolTestCase, caplog):
diff --git a/backend/evaluation_tests/collect_experiences_agent/memory_hallucination_baseline_test.py b/backend/evaluation_tests/collect_experiences_agent/memory_hallucination_baseline_test.py
index bcab82a09..ad0821300 100644
--- a/backend/evaluation_tests/collect_experiences_agent/memory_hallucination_baseline_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/memory_hallucination_baseline_test.py
@@ -108,7 +108,7 @@ async def collect_experiences_exec(caplog: LogCaptureFixture, test_case: Scripte
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(5)
 async def test_memory_recall_accuracy_after_summarization(
         caplog: LogCaptureFixture,
diff --git a/backend/evaluation_tests/collect_experiences_agent/recap_field_accuracy_test.py b/backend/evaluation_tests/collect_experiences_agent/recap_field_accuracy_test.py
index 12634241c..bcec6e644 100644
--- a/backend/evaluation_tests/collect_experiences_agent/recap_field_accuracy_test.py
+++ b/backend/evaluation_tests/collect_experiences_agent/recap_field_accuracy_test.py
@@ -104,7 +104,7 @@ async def collect_experiences_exec(caplog: LogCaptureFixture, test_case: Scripte
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_recap_contains_only_structured_fields(
         caplog: LogCaptureFixture,
diff --git a/backend/evaluation_tests/cv_parser/test_cv_parser.py b/backend/evaluation_tests/cv_parser/test_cv_parser.py
index c651c75b1..b039c04eb 100644
--- a/backend/evaluation_tests/cv_parser/test_cv_parser.py
+++ b/backend/evaluation_tests/cv_parser/test_cv_parser.py
@@ -35,7 +35,7 @@ def _to_markdown(self) -> str:
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize("case", test_cases_to_run, ids=[c.name for c in test_cases_to_run])
 async def test_cv_parser(case: CVParserTestCase, common_folder_path: str):
diff --git a/backend/evaluation_tests/linking_and_ranking_pipeline/experience_pipeline_test.py b/backend/evaluation_tests/linking_and_ranking_pipeline/experience_pipeline_test.py
index 9a6bb8b04..4ec4bed42 100644
--- a/backend/evaluation_tests/linking_and_ranking_pipeline/experience_pipeline_test.py
+++ b/backend/evaluation_tests/linking_and_ranking_pipeline/experience_pipeline_test.py
@@ -275,7 +275,7 @@ class ExperiencePipelineTestCase(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize("test_case", get_test_cases_to_run(test_cases), ids=[test_case.name for test_case in get_test_cases_to_run(test_cases)])
 async def test_experience_pipeline(test_case: ExperiencePipelineTestCase, setup_search_services: Awaitable[SearchServices], caplog: pytest.LogCaptureFixture):
diff --git a/backend/evaluation_tests/linking_and_ranking_pipeline/infer_occupation_tool/_contextualization_llm_test.py b/backend/evaluation_tests/linking_and_ranking_pipeline/infer_occupation_tool/_contextualization_llm_test.py
index 5dca4fd8f..526cd5a27 100644
--- a/backend/evaluation_tests/linking_and_ranking_pipeline/infer_occupation_tool/_contextualization_llm_test.py
+++ b/backend/evaluation_tests/linking_and_ranking_pipeline/infer_occupation_tool/_contextualization_llm_test.py
@@ -94,7 +94,7 @@ class ContextualizationTestCase(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.parametrize("test_case", get_test_cases_to_run(test_cases), ids=[test_case.name for test_case in get_test_cases_to_run(test_cases)])
 async def test_relevant_occupations_classifier_llm(test_case: ContextualizationTestCase, caplog: LogCaptureFixture):
     logger = logging.getLogger(__name__)
diff --git a/backend/evaluation_tests/linking_and_ranking_pipeline/pick_top_skills_tool_test.py b/backend/evaluation_tests/linking_and_ranking_pipeline/pick_top_skills_tool_test.py
index e4a4b2770..669804cb9 100644
--- a/backend/evaluation_tests/linking_and_ranking_pipeline/pick_top_skills_tool_test.py
+++ b/backend/evaluation_tests/linking_and_ranking_pipeline/pick_top_skills_tool_test.py
@@ -133,7 +133,7 @@ class PickTopSkillToolTestCase(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize("test_case", get_test_cases_to_run(test_cases), ids=[test_case.name for test_case in get_test_cases_to_run(test_cases)])
 async def test_pick_top_skill_tool(test_case: PickTopSkillToolTestCase, caplog):
diff --git a/backend/evaluation_tests/skill_explorer_agent/_conversation_llm_test.py b/backend/evaluation_tests/skill_explorer_agent/_conversation_llm_test.py
index 1924c7968..30592e06a 100644
--- a/backend/evaluation_tests/skill_explorer_agent/_conversation_llm_test.py
+++ b/backend/evaluation_tests/skill_explorer_agent/_conversation_llm_test.py
@@ -192,7 +192,7 @@ class _TestCaseConversation(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
diff --git a/backend/evaluation_tests/skill_explorer_agent/_responsibilities_extraction_tool_test.py b/backend/evaluation_tests/skill_explorer_agent/_responsibilities_extraction_tool_test.py
index 4fcea4cf8..9e9059452 100644
--- a/backend/evaluation_tests/skill_explorer_agent/_responsibilities_extraction_tool_test.py
+++ b/backend/evaluation_tests/skill_explorer_agent/_responsibilities_extraction_tool_test.py
@@ -145,7 +145,7 @@ class _TestCaseDataExtraction(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases_data_extraction),
                          ids=[case.name for case in get_test_cases_to_run(test_cases_data_extraction)])
diff --git a/backend/evaluation_tests/skill_explorer_agent/_sentence_decomposition_llm_test.py b/backend/evaluation_tests/skill_explorer_agent/_sentence_decomposition_llm_test.py
index 09057cd97..7d6cb11c2 100644
--- a/backend/evaluation_tests/skill_explorer_agent/_sentence_decomposition_llm_test.py
+++ b/backend/evaluation_tests/skill_explorer_agent/_sentence_decomposition_llm_test.py
@@ -129,7 +129,7 @@ class _TestCaseSentenceDecomposition(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases_sentence_decomposition),
                          ids=[case.name for case in get_test_cases_to_run(test_cases_sentence_decomposition)])
diff --git a/backend/evaluation_tests/skill_explorer_agent/loop_detection_scripted_user_test.py b/backend/evaluation_tests/skill_explorer_agent/loop_detection_scripted_user_test.py
index fc96994b7..9f5f6700d 100644
--- a/backend/evaluation_tests/skill_explorer_agent/loop_detection_scripted_user_test.py
+++ b/backend/evaluation_tests/skill_explorer_agent/loop_detection_scripted_user_test.py
@@ -168,7 +168,7 @@ class ScriptedLoopDetectionTestCase(ScriptedUserEvaluationTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize(
     "test_case",
diff --git a/backend/evaluation_tests/skill_explorer_agent/loop_detection_test.py b/backend/evaluation_tests/skill_explorer_agent/loop_detection_test.py
index d3285b84b..b6e0bb340 100644
--- a/backend/evaluation_tests/skill_explorer_agent/loop_detection_test.py
+++ b/backend/evaluation_tests/skill_explorer_agent/loop_detection_test.py
@@ -230,7 +230,7 @@ class LoopDetectionTestCase(EvaluationTestCase):
 ]
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize(
     "test_case",
diff --git a/backend/evaluation_tests/skill_explorer_agent/skills_explorer_agent_simulated_user_test.py b/backend/evaluation_tests/skill_explorer_agent/skills_explorer_agent_simulated_user_test.py
index 91b3f5dec..06a675e35 100644
--- a/backend/evaluation_tests/skill_explorer_agent/skills_explorer_agent_simulated_user_test.py
+++ b/backend/evaluation_tests/skill_explorer_agent/skills_explorer_agent_simulated_user_test.py
@@ -22,7 +22,7 @@
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
diff --git a/backend/evaluation_tests/summarizer/summarizer_test.py b/backend/evaluation_tests/summarizer/summarizer_test.py
index 689f30676..f40ce0eea 100644
--- a/backend/evaluation_tests/summarizer/summarizer_test.py
+++ b/backend/evaluation_tests/summarizer/summarizer_test.py
@@ -17,7 +17,7 @@ async def _evaluate_with_llm(prompt: str) -> str:
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_summarizer_without_existing_summary(fake_conversation_context: FakeConversationContext,
                                                    common_folder_path: str):
@@ -70,7 +70,7 @@ async def test_summarizer_without_existing_summary(fake_conversation_context: Fa
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 async def test_summarizer_with_existing_summary(fake_conversation_context: FakeConversationContext,
                                                 common_folder_path: str):
diff --git a/backend/evaluation_tests/translation_tool_test.py b/backend/evaluation_tests/translation_tool_test.py
index f30735372..13ffb74b1 100644
--- a/backend/evaluation_tests/translation_tool_test.py
+++ b/backend/evaluation_tests/translation_tool_test.py
@@ -42,7 +42,7 @@ class _TestCase(CompassTestCase):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('current_test_case', get_test_cases_to_run(test_cases),
                          ids=[case.test_id for case in get_test_cases_to_run(test_cases)])
diff --git a/backend/evaluation_tests/welcome_agent/welcome_agent_no_redundant_questions_test.py b/backend/evaluation_tests/welcome_agent/welcome_agent_no_redundant_questions_test.py
index 91096ee63..139a74f8e 100644
--- a/backend/evaluation_tests/welcome_agent/welcome_agent_no_redundant_questions_test.py
+++ b/backend/evaluation_tests/welcome_agent/welcome_agent_no_redundant_questions_test.py
@@ -89,7 +89,7 @@ def _assert_no_redundant_questions(response_text: str):
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(5)
 async def test_welcome_agent_does_not_ask_about_known_personal_data(evals_setup, caplog):
     """
diff --git a/backend/evaluation_tests/welcome_agent/welcome_agent_scripted_user_test.py b/backend/evaluation_tests/welcome_agent/welcome_agent_scripted_user_test.py
index 5af42c975..b36041bfd 100644
--- a/backend/evaluation_tests/welcome_agent/welcome_agent_scripted_user_test.py
+++ b/backend/evaluation_tests/welcome_agent/welcome_agent_scripted_user_test.py
@@ -167,7 +167,7 @@ def event_loop():
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
diff --git a/backend/evaluation_tests/welcome_agent/welcome_agent_simulated_user_test.py b/backend/evaluation_tests/welcome_agent/welcome_agent_simulated_user_test.py
index c07909bc5..e25f01727 100644
--- a/backend/evaluation_tests/welcome_agent/welcome_agent_simulated_user_test.py
+++ b/backend/evaluation_tests/welcome_agent/welcome_agent_simulated_user_test.py
@@ -32,7 +32,7 @@ def event_loop():
 
 
 @pytest.mark.asyncio
-@pytest.mark.evaluation_test("gemini-2.5-flash-lite/")
+@pytest.mark.evaluation_test("gemini-3.5-flash-lite/")
 @pytest.mark.repeat(3)
 @pytest.mark.parametrize('test_case', get_test_cases_to_run(test_cases),
                          ids=[case.name for case in get_test_cases_to_run(test_cases)])
diff --git a/iac/backend/deploy_backend.py b/iac/backend/deploy_backend.py
index 7b13b70ec..901049b77 100644
--- a/iac/backend/deploy_backend.py
+++ b/iac/backend/deploy_backend.py
@@ -54,7 +54,7 @@ class BackendServiceConfig:
     langfuse_public_key: Optional[str]
     langfuse_secret_key: Optional[str | pulumi.Output[str]]
     tracing_config: Optional[str]
-    # LLM provider and per-tier models. All optional; unset deploys on the Gemini 2.5 defaults.
+    # LLM provider and per-tier models. All optional; unset deploys on the Gemini defaults.
     # The Anthropic API key must be provisioned as a secret, never as a plain environment value.
     llm_provider: Optional[str]
     llm_default_model: Optional[str]
diff --git a/iac/templates/env.template b/iac/templates/env.template
index 32a2fc146..5802db5da 100644
--- a/iac/templates/env.template
+++ b/iac/templates/env.template
@@ -53,8 +53,8 @@ BACKEND_LANGFUSE_PUBLIC_KEY=<BACKEND_LANGFUSE_PUBLIC_KEY>
 BACKEND_LANGFUSE_SECRET_KEY=<BACKEND_LANGFUSE_SECRET_KEY>
 BACKEND_TRACING_CONFIG=<JSON_STRING>
 
-# LLM provider and per-tier model settings - all optional, leave empty to deploy on the Gemini 2.5 defaults
-# (gemini-2.5-flash-lite / gemini-2.5-flash / gemini-2.5-pro).
+# LLM provider and per-tier model settings - all optional, leave empty to deploy on the Gemini defaults
+# (gemini-3.5-flash-lite / gemini-3.8-flash / gemini-3.1-pro-preview). The Gemini 3.x models require VERTEX_API_GEN_AI_REGION=global.
 # For a non-Gemini provider, an empty tier falls back to the next tier down that is set
 # (deep reasoning -> reasoning -> default). See backend/.env.example for details.
 LLM_PROVIDER=/.*/