diff --git a/CLAUDE.md b/CLAUDE.md
index 0d3f517..3b0cf16 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -311,7 +311,7 @@ owner = "frappe"         # Default repository owner (optional)
 
 [openai]
 api_key = "sk-xxxxx"     # OpenAI API key (required)
-model = "gpt-4.1"        # Model to use (default: "gpt-4.1")
+model = "openai:gpt-4.1" # Model to use (default: "openai:gpt-4.1")
 max_patch_size = 10000   # Max patch size before fallback (default: 10000)
 
 [database]
diff --git a/README.md b/README.md
index 9efdf3a..dde0d98 100644
--- a/README.md
+++ b/README.md
@@ -62,7 +62,7 @@ nano ~/.pretty-release-notes/config.toml
 
 ### Configuration Format
 
-The configuration file uses TOML format with sections for GitHub credentials, LLM settings, database caching, and filters. The canonical section name is `[llm]`, while the legacy `[openai]` section is still accepted for backward compatibility. Plain model names still default to OpenAI, and `provider:model` targets other `any-llm` providers. See [`config.toml.example`](config.toml.example) for the complete structure and all available options.
+The configuration file uses TOML format with sections for GitHub credentials, LLM settings, database caching, and filters. The canonical section name is `[llm]`, while the legacy `[openai]` section is still accepted for backward compatibility. Prefer fully qualified `provider:model` values such as `openai:gpt-4.1`; unqualified model names are still accepted and default to OpenAI for backward compatibility. See [`config.toml.example`](config.toml.example) for the complete structure and all available options.
 
 You can override the config location using the `--config-path` flag.
 
@@ -130,7 +130,7 @@ from pretty_release_notes import ReleaseNotesBuilder
 client = (
     ReleaseNotesBuilder()
     .with_github_token("ghp_your_token")
-    .with_llm("sk_your_key", model="gpt-4")  # or model="anthropic:claude-sonnet-4-5"
+    .with_llm("sk_your_key", model="openai:gpt-4.1")  # or model="anthropic:claude-sonnet-4-5"
     .with_database("sqlite", enabled=True)
     .with_filters(
         exclude_types={"chore", "ci", "refactor"},
@@ -197,7 +197,7 @@ curl -X POST http://localhost:8000/generate \
     "previous_tag_name": "v15.38.0",
     "github_token": "ghp_your_token_here",
     "llm_key": "sk-your_key_here",
-    "llm_model": "gpt-4",
+    "llm_model": "openai:gpt-4.1",
     "exclude_types": ["chore", "ci", "refactor"],
     "exclude_labels": ["skip-release-notes"],
     "exclude_authors": ["dependabot[bot]"]
diff --git a/config.toml.example b/config.toml.example
index 385da03..c5d8dd3 100644
--- a/config.toml.example
+++ b/config.toml.example
@@ -19,13 +19,13 @@ owner = "frappe"
 
 [llm]
 # LLM provider API key (required)
-# For plain OpenAI models, use an OpenAI key. For other providers, use that provider's key.
+# Use the API key for the provider referenced by `model`.
 # Legacy [openai] is still accepted for backward compatibility.
 api_key = ""
 
-# Model to use. Plain names default to OpenAI.
-# For other providers, use "provider:model" (for example: "anthropic:claude-sonnet-4-5")
-model = "gpt-4.1"
+# Model to use. Prefer fully qualified "provider:model" values.
+# Unqualified names still default to OpenAI for backward compatibility.
+model = "openai:gpt-4.1"
 
 # Maximum patch size before fallback to commit message (default: 10000)
 max_patch_size = 10000
diff --git a/docs/adr/003-toml-configuration.md b/docs/adr/003-toml-configuration.md
index 865ee28..5ccfde3 100644
--- a/docs/adr/003-toml-configuration.md
+++ b/docs/adr/003-toml-configuration.md
@@ -40,9 +40,9 @@ EXCLUDE_PR_TYPES=chore,refactor,ci
 token = "ghp_xxxxx"
 owner = "frappe"
 
-[openai]
+[llm]
 api_key = "sk-xxxxx"
-model = "gpt-4.1"
+model = "openai:gpt-4.1"
 
 [filters]
 exclude_change_types = ["chore", "refactor", "ci"]
@@ -100,9 +100,9 @@ force_use_commits = false # Force using commits over PRs
 token = ""        # Required
 owner = ""        # Optional default owner
 
-[openai]          # OpenAI API settings
+[llm]          # LLM API settings
 api_key = ""      # Required
-model = ""        # Optional, default: "gpt-4.1"
+model = ""        # Optional, default: "openai:gpt-4.1"
 max_patch_size    # Optional, default: 10000
 
 [database]        # Cache configuration
diff --git a/examples/library_usage.py b/examples/library_usage.py
index 2396d77..3e74596 100644
--- a/examples/library_usage.py
+++ b/examples/library_usage.py
@@ -37,7 +37,7 @@ def advanced_usage():
 	client = (
 		ReleaseNotesBuilder()
 		.with_github_token("ghp_xxxxx")  # Replace with your token
-		.with_llm("sk-xxxxx", model="gpt-4", max_patch_size=15000)
+		.with_llm("sk-xxxxx", model="openai:gpt-4.1", max_patch_size=15000)
 		.with_database("sqlite", enabled=True)
 		.with_filters(
 			exclude_types={"chore", "refactor", "ci", "style", "test"},
@@ -76,7 +76,7 @@ def direct_config_usage():
 
 	config = ReleaseNotesConfig(
 		github=GitHubConfig(token="ghp_xxxxx"),  # Replace with your token
-		llm=LLMConfig(api_key="sk-xxxxx", model="gpt-4.1"),  # Replace with your key
+		llm=LLMConfig(api_key="sk-xxxxx", model="openai:gpt-4.1"),  # Replace with your key
 		database=DatabaseConfig(type="sqlite", enabled=True),
 		filters=FilterConfig(
 			exclude_change_types={"chore", "refactor"},
diff --git a/pretty_release_notes/api.py b/pretty_release_notes/api.py
index 5864557..600018c 100644
--- a/pretty_release_notes/api.py
+++ b/pretty_release_notes/api.py
@@ -12,6 +12,7 @@
 )
 from .core.interfaces import NullProgressReporter, ProgressReporter
 from .generator import ReleaseNotesGenerator
+from .openai_client import DEFAULT_MODEL
 
 
 class ReleaseNotesClient:
@@ -73,7 +74,7 @@ class ReleaseNotesBuilder:
 	def __init__(self):
 		self._github_token = None
 		self._llm_key = None
-		self._llm_model = "gpt-4.1"
+		self._llm_model = DEFAULT_MODEL
 		self._max_patch_size = 10000
 		self._db_type = "sqlite"
 		self._db_name = "stored_lines"
@@ -93,14 +94,16 @@ def with_github_token(self, token: str) -> "ReleaseNotesBuilder":
 		self._github_token = token
 		return self
 
-	def with_llm(self, api_key: str, model: str = "gpt-4.1", max_patch_size: int = 10000) -> "ReleaseNotesBuilder":
+	def with_llm(self, api_key: str, model: str = DEFAULT_MODEL, max_patch_size: int = 10000) -> "ReleaseNotesBuilder":
 		"""Set LLM configuration."""
 		self._llm_key = api_key
 		self._llm_model = model
 		self._max_patch_size = max_patch_size
 		return self
 
-	def with_openai(self, api_key: str, model: str = "gpt-4.1", max_patch_size: int = 10000) -> "ReleaseNotesBuilder":
+	def with_openai(
+		self, api_key: str, model: str = DEFAULT_MODEL, max_patch_size: int = 10000
+	) -> "ReleaseNotesBuilder":
 		"""Backward-compatible alias for with_llm()."""
 		return self.with_llm(api_key=api_key, model=model, max_patch_size=max_patch_size)
 
diff --git a/pretty_release_notes/core/config.py b/pretty_release_notes/core/config.py
index e89ade1..31f1545 100644
--- a/pretty_release_notes/core/config.py
+++ b/pretty_release_notes/core/config.py
@@ -1,6 +1,8 @@
 from dataclasses import dataclass, field
 from pathlib import Path
 
+from ..openai_client import DEFAULT_MODEL
+
 
 @dataclass
 class GitHubConfig:
@@ -15,7 +17,7 @@ def __post_init__(self):
 @dataclass
 class LLMConfig:
 	api_key: str
-	model: str = "gpt-4.1"
+	model: str = DEFAULT_MODEL
 	max_patch_size: int = 10000
 
 	def __post_init__(self):
diff --git a/pretty_release_notes/core/config_loader.py b/pretty_release_notes/core/config_loader.py
index 9e022b7..bc6bd99 100644
--- a/pretty_release_notes/core/config_loader.py
+++ b/pretty_release_notes/core/config_loader.py
@@ -5,6 +5,7 @@
 
 from dotenv import dotenv_values
 
+from ..openai_client import DEFAULT_MODEL
 from .config import (
 	DatabaseConfig,
 	FilterConfig,
@@ -41,7 +42,7 @@ def load(self) -> ReleaseNotesConfig:
 			),
 			llm=LLMConfig(
 				api_key=llm_api_key,
-				model=self.config_dict.get("llm_model", self.config_dict.get("openai_model", "gpt-4.1")),
+				model=self.config_dict.get("llm_model", self.config_dict.get("openai_model", DEFAULT_MODEL)),
 				max_patch_size=self.config_dict.get("max_patch_size", 10000),
 			),
 			database=DatabaseConfig(
@@ -97,7 +98,7 @@ def load(self) -> ReleaseNotesConfig:
 			github=GitHubConfig(token=github_token, owner=config.get("DEFAULT_OWNER")),
 			llm=LLMConfig(
 				api_key=llm_key,
-				model=config.get("LLM_MODEL") or config.get("OPENAI_MODEL") or "gpt-4.1",
+				model=config.get("LLM_MODEL") or config.get("OPENAI_MODEL") or DEFAULT_MODEL,
 				max_patch_size=int(config.get("MAX_PATCH_SIZE") or "10000"),
 			),
 			database=DatabaseConfig(
@@ -180,7 +181,7 @@ def load(self) -> ReleaseNotesConfig:
 			),
 			llm=LLMConfig(
 				api_key=llm_key,
-				model=llm_config.get("model", "gpt-4.1"),
+				model=llm_config.get("model", DEFAULT_MODEL),
 				max_patch_size=llm_config.get("max_patch_size", 10000),
 			),
 			database=DatabaseConfig(
diff --git a/pretty_release_notes/openai_client.py b/pretty_release_notes/openai_client.py
index 10c1135..0a1fb8f 100644
--- a/pretty_release_notes/openai_client.py
+++ b/pretty_release_notes/openai_client.py
@@ -8,6 +8,7 @@
 )
 
 DEFAULT_PROVIDER = "openai"
+DEFAULT_MODEL = "openai:gpt-4.1"
 OPENAI_MODELS_WITH_FLEX = {
 	"o3",
 	"o4-mini",
diff --git a/pretty_release_notes/setup_command.py b/pretty_release_notes/setup_command.py
index 5e39853..0a855db 100644
--- a/pretty_release_notes/setup_command.py
+++ b/pretty_release_notes/setup_command.py
@@ -7,6 +7,8 @@
 from rich.console import Console
 from rich.prompt import Confirm, Prompt
 
+from .openai_client import DEFAULT_MODEL
+
 console = Console()
 
 
@@ -80,8 +82,8 @@ def setup_config(
 		password=True,
 	)
 	llm_model = Prompt.ask(
-		"Model (use provider:model for non-OpenAI providers)",
-		default=existing_values.get("llm_model") or "gpt-4.1",
+		"Model (prefer provider:model, e.g. openai:gpt-4.1)",
+		default=existing_values.get("llm_model") or DEFAULT_MODEL,
 	)
 	max_patch_size = Prompt.ask(
 		"Maximum patch size before fallback",
@@ -259,7 +261,8 @@ def to_toml_array(s: str) -> str:
 [llm]
 # Legacy [openai] is still accepted for backward compatibility.
 api_key = "{llm_key}"
-# Use plain model names for OpenAI or "provider:model" for other providers.
+# Prefer "provider:model" syntax (for example: "openai:gpt-4.1" or "anthropic:claude-sonnet-4-5").
+# Unqualified names are still treated as OpenAI for backward compatibility.
 model = "{llm_model}"
 max_patch_size = {max_patch_size}
 
diff --git a/pretty_release_notes/web/app.py b/pretty_release_notes/web/app.py
index e580901..e6cf275 100644
--- a/pretty_release_notes/web/app.py
+++ b/pretty_release_notes/web/app.py
@@ -8,6 +8,7 @@
 
 from ..api import ReleaseNotesBuilder
 from ..core.interfaces import ProgressEvent, ProgressReporter
+from ..openai_client import DEFAULT_MODEL
 
 app = FastAPI(title="Pretty Release Notes API", version="1.0.0")
 
@@ -25,7 +26,7 @@ class GenerateRequest(BaseModel):
 	tag: str
 	github_token: str
 	llm_key: str = Field(validation_alias=AliasChoices("llm_key", "openai_key"))
-	llm_model: str = Field(default="gpt-4.1", validation_alias=AliasChoices("llm_model", "openai_model"))
+	llm_model: str = Field(default=DEFAULT_MODEL, validation_alias=AliasChoices("llm_model", "openai_model"))
 	exclude_types: list[str] = []
 	exclude_labels: list[str] = []
 	exclude_authors: list[str] = []
diff --git a/tests/test_api.py b/tests/test_api.py
index c5fb79b..0e7ecf5 100644
--- a/tests/test_api.py
+++ b/tests/test_api.py
@@ -13,6 +13,7 @@
 	ReleaseNotesConfig,
 )
 from pretty_release_notes.core.interfaces import ProgressEvent, ProgressReporter
+from pretty_release_notes.openai_client import DEFAULT_MODEL
 
 
 class TestProgressReporting:
@@ -222,7 +223,7 @@ def test_llm_defaults(self):
 		client = ReleaseNotesBuilder().with_github_token("test_token").with_llm("test_key").build()
 
 		# Check defaults
-		assert client.config.llm.model == "gpt-4.1"
+		assert client.config.llm.model == DEFAULT_MODEL
 		assert client.config.llm.max_patch_size == 10000
 
 
diff --git a/tests/test_core.py b/tests/test_core.py
index 19012fb..e6ff64b 100644
--- a/tests/test_core.py
+++ b/tests/test_core.py
@@ -20,6 +20,7 @@
 	ProgressEvent,
 	ProgressReporter,
 )
+from pretty_release_notes.openai_client import DEFAULT_MODEL
 
 
 class TestProgressEvent:
@@ -95,7 +96,7 @@ class TestLLMConfig:
 	def test_valid_config_with_defaults(self):
 		config = LLMConfig(api_key="test_key")
 		assert config.api_key == "test_key"
-		assert config.model == "gpt-4.1"
+		assert config.model == DEFAULT_MODEL
 		assert config.max_patch_size == 10000
 
 	def test_valid_config_with_custom_values(self):
diff --git a/tests/test_openai_client.py b/tests/test_openai_client.py
index d26647e..e7b14ec 100644
--- a/tests/test_openai_client.py
+++ b/tests/test_openai_client.py
@@ -3,7 +3,7 @@
 from types import SimpleNamespace
 from unittest.mock import patch
 
-from pretty_release_notes.openai_client import format_model_name, get_chat_response
+from pretty_release_notes.openai_client import DEFAULT_MODEL, format_model_name, get_chat_response
 
 
 def _mock_completion_response(content: str):
@@ -16,7 +16,26 @@ class TestOpenAIClient:
 	"""Test the compatibility wrapper around any-llm."""
 
 	@patch("pretty_release_notes.openai_client.completion")
-	def test_plain_model_defaults_to_openai_provider(self, mock_completion):
+	def test_default_model_is_passed_through_to_any_llm(self, mock_completion):
+		mock_completion.return_value = _mock_completion_response("summary")
+
+		result = get_chat_response(
+			content="Write a summary",
+			model=DEFAULT_MODEL,
+			api_key="test-key",
+		)
+
+		assert result == "summary"
+		mock_completion.assert_called_once_with(
+			messages=[{"role": "user", "content": "Write a summary"}],
+			model=DEFAULT_MODEL,
+			api_key="test-key",
+			client_args={"timeout": 900.0},
+			service_tier="auto",
+		)
+
+	@patch("pretty_release_notes.openai_client.completion")
+	def test_plain_model_defaults_to_openai_provider_for_backward_compatibility(self, mock_completion):
 		mock_completion.return_value = _mock_completion_response("summary")
 
 		result = get_chat_response(
@@ -72,6 +91,7 @@ def test_provider_prefixed_model_is_passed_through_to_any_llm(self, mock_complet
 		)
 
 	def test_format_model_name_supports_provider_prefixed_models(self):
+		assert format_model_name("openai:gpt-4.1") == "OpenAI gpt-4.1"
 		assert format_model_name("gpt-4.1") == "OpenAI gpt-4.1"
 		assert format_model_name("openrouter:deepseek-r1") == "openrouter:deepseek-r1"