feat(llm): refresh the model lists to each provider's current lineup

- xAI: Grok 4.6 and 4.5; DeepSeek: deepseek-flash (V4.1); Qwen: the 3.8 series; Mistral: the current dated snapshots
- every provider can name a model the picker does not list
- retired IDs stay valid, so existing configs run without a warning

tests/test_model_validation.py

@@ -64,3 +64,37 @@ def test_legacy_ids_stay_valid_without_being_offered():
         for model in ids:
             assert validate_model(provider, model), model
             assert model not in offered, f"{model} is legacy but still in the picker"
+
+
+@pytest.mark.unit
+def test_an_explicit_alias_of_a_listed_model_is_known():
+    """gpt-5.6 is served under its own name and as gpt-5.6-sol; naming the
+    explicit one should not warn that the model is unknown."""
+    from tradingagents.llm_clients.validators import validate_model
+
+    assert validate_model("openai", "gpt-5.6-sol")
+
+
+@pytest.mark.unit
+@pytest.mark.parametrize("provider", ["openai", "anthropic", "google", "xai"])
+@pytest.mark.parametrize("mode", ["quick", "deep"])
+def test_every_provider_lets_you_name_your_own_model(provider, mode):
+    """The docs tell users to name any model their provider serves; the picker
+    has to offer that too, or a new model is unreachable until we ship a list."""
+    from tradingagents.llm_clients.model_catalog import get_model_options
+
+    assert "custom" in [value for _, value in get_model_options(provider, mode)]
+
+
+@pytest.mark.unit
+@pytest.mark.parametrize("provider, model", [
+    ("xai", "grok-4.20-0309-reasoning"),
+    ("deepseek", "deepseek-v4-flash"),
+    ("qwen", "qwen3.7-max"),
+])
+def test_a_retired_model_id_still_runs_without_a_warning(provider, model):
+    """A config written against an earlier release keeps working: the provider
+    still serves these, they are just no longer offered in the picker."""
+    from tradingagents.llm_clients.validators import validate_model
+
+    assert validate_model(provider, model)