diff --git a/openhands_cli/tui/modals/settings/choices.py b/openhands_cli/tui/modals/settings/choices.py
index 231924eb..3756de9a 100644
--- a/openhands_cli/tui/modals/settings/choices.py
+++ b/openhands_cli/tui/modals/settings/choices.py
@@ -1,20 +1,49 @@
+import litellm
+
 from openhands.sdk.llm import UNVERIFIED_MODELS_EXCLUDING_BEDROCK, VERIFIED_MODELS
 
 
+# Get set of valid litellm provider names for filtering
+# See: https://docs.litellm.ai/docs/providers
+_VALID_LITELLM_PROVIDERS: set[str] = {
+    str(getattr(p, "value", p)) for p in litellm.provider_list
+}
+
+
 def get_provider_options() -> list[tuple[str, str]]:
-    """Get list of available LLM providers."""
-    providers = list(VERIFIED_MODELS.keys()) + list(
-        UNVERIFIED_MODELS_EXCLUDING_BEDROCK.keys()
-    )
-    return [(provider, provider) for provider in providers]
+    """Get list of available LLM providers.
+
+    Includes:
+    - All VERIFIED_MODELS providers (openhands, openai, anthropic, mistral)
+      even if not in litellm.provider_list (e.g. 'openhands' is custom)
+    - UNVERIFIED providers that are known to litellm (filters out invalid
+      "providers" like 'meta-llama', 'Qwen' which are vendor names)
+
+    Sorted alphabetically.
+    """
+    # Verified providers always included (includes custom like 'openhands')
+    verified_providers = set(VERIFIED_MODELS.keys())
+
+    # Unverified providers are filtered to only valid litellm providers
+    unverified_providers = set(UNVERIFIED_MODELS_EXCLUDING_BEDROCK.keys())
+    valid_unverified = unverified_providers & _VALID_LITELLM_PROVIDERS
+
+    # Combine and sort
+    all_valid_providers = sorted(verified_providers | valid_unverified)
+
+    return [(provider, provider) for provider in all_valid_providers]
 
 
 def get_model_options(provider: str) -> list[tuple[str, str]]:
-    """Get list of available models for a provider."""
+    """Get list of available models for a provider, sorted alphabetically."""
     models = VERIFIED_MODELS.get(
         provider, []
     ) + UNVERIFIED_MODELS_EXCLUDING_BEDROCK.get(provider, [])
-    return [(model, model) for model in models]
+
+    # Remove duplicates and sort
+    unique_models = sorted(set(models))
+
+    return [(model, model) for model in unique_models]
 
 
 provider_options = get_provider_options()
diff --git a/openhands_cli/tui/modals/settings/utils.py b/openhands_cli/tui/modals/settings/utils.py
index 7c7d761a..5fa36641 100644
--- a/openhands_cli/tui/modals/settings/utils.py
+++ b/openhands_cli/tui/modals/settings/utils.py
@@ -79,10 +79,11 @@ def get_full_model_name(self):
             return str(self.custom_model)
 
         model_str = str(self.model)
-        full_model = (
-            f"{self.provider}/{model_str}" if "/" not in model_str else model_str
-        )
-        return full_model
+
+        # Always add provider prefix - litellm requires it for routing.
+        # Even if model contains '/' (e.g. "openai/gpt-4.1" from openrouter)
+        # See: https://docs.litellm.ai/docs/providers
+        return f"{self.provider}/{model_str}"
 
 
 class SettingsSaveResult(BaseModel):
diff --git a/tests/tui/modals/settings/test_settings_utils.py b/tests/tui/modals/settings/test_settings_utils.py
index 93fa1173..c870bfb6 100644
--- a/tests/tui/modals/settings/test_settings_utils.py
+++ b/tests/tui/modals/settings/test_settings_utils.py
@@ -132,6 +132,40 @@ def test_preserves_existing_api_key_when_not_provided(
             None,  # advanced base_url cleared
             ("custom_model", "base_url"),
         ),
+        (
+            "basic",
+            "openrouter",
+            "google/gemini-3-flash-preview",
+            "should-be-cleared",
+            "https://advanced.example",
+            # All providers require prefix even for models with '/' in their name
+            # See: https://docs.litellm.ai/docs/providers/openrouter
+            "openrouter/google/gemini-3-flash-preview",
+            None,
+            ("custom_model", "base_url"),
+        ),
+        (
+            "basic",
+            "nvidia_nim",
+            "meta/llama3-70b-instruct",
+            "should-be-cleared",
+            "https://advanced.example",
+            # See: https://docs.litellm.ai/docs/providers/nvidia_nim
+            "nvidia_nim/meta/llama3-70b-instruct",
+            None,
+            ("custom_model", "base_url"),
+        ),
+        (
+            "basic",
+            "deepinfra",
+            "meta-llama/Meta-Llama-3.1-8B-Instruct",
+            "should-be-cleared",
+            "https://advanced.example",
+            # See: https://docs.litellm.ai/docs/providers
+            "deepinfra/meta-llama/Meta-Llama-3.1-8B-Instruct",
+            None,
+            ("custom_model", "base_url"),
+        ),
         (
             "advanced",
             "openai",