diff --git a/navi/api/deps.py b/navi/api/deps.py
index 9aed576..0881e00 100644
--- a/navi/api/deps.py
+++ b/navi/api/deps.py
@@ -18,20 +18,20 @@
 
 
 @lru_cache
-def _registries() -> tuple[ToolRegistry, ProfileRegistry, BackendRegistry]:
+def get_registries() -> tuple[ToolRegistry, ProfileRegistry, BackendRegistry]:
     return build_default_registries()
 
 
 def get_tool_registry() -> ToolRegistry:
-    return _registries()[0]
+    return get_registries()[0]
 
 
 def get_profile_registry() -> ProfileRegistry:
-    return _registries()[1]
+    return get_registries()[1]
 
 
 def get_backend_registry() -> BackendRegistry:
-    return _registries()[2]
+    return get_registries()[2]
 
 
 _session_store = SqliteSessionStore(settings.db_path)
diff --git a/navi/api/websocket.py b/navi/api/websocket.py
index e00d272..3a5b1cb 100644
--- a/navi/api/websocket.py
+++ b/navi/api/websocket.py
@@ -41,8 +41,8 @@
     log.info("ws.connected", session_id=session_id)
 
     # Build agent (can't use FastAPI Depends inside WebSocket directly)
-    from navi.api.deps import _registries
-    tools, profiles, backends = _registries()
+    from navi.api.deps import get_registries
+    tools, profiles, backends = get_registries()
     agent = Agent(session_store, profiles, tools, backends)
 
     try:
diff --git a/navi/core/agent.py b/navi/core/agent.py
index ef36fa7..d5af9f1 100644
--- a/navi/core/agent.py
+++ b/navi/core/agent.py
@@ -108,7 +108,7 @@
         profile = self._profiles.get(session.profile_id)
         tools = self._tool_list(profile.enabled_tools)
         tool_schemas = [t.schema() for t in tools]
-        llm = self._get_backend(profile.llm_backend, profile.model)
+        llm = self._get_backend(profile.llm_backend)
 
         # Inject system prompt on first message
         if not session.messages:
@@ -122,6 +122,7 @@
                 session.messages,
                 tools=tool_schemas if tools else None,
                 temperature=profile.temperature,
+                model=profile.model,
             )
 
             if response.finish_reason == "stop" or not response.tool_calls:
@@ -161,7 +162,7 @@
         profile = self._profiles.get(session.profile_id)
         tools = self._tool_list(profile.enabled_tools)
         tool_schemas = [t.schema() for t in tools]
-        llm = self._get_backend(profile.llm_backend, profile.model)
+        llm = self._get_backend(profile.llm_backend)
 
         if not session.messages:
             session.messages.append(Message(role="system", content=self._build_system_prompt(profile.system_prompt)))
@@ -174,6 +175,7 @@
                 session.messages,
                 tools=tool_schemas if tools else None,
                 temperature=profile.temperature,
+                model=profile.model,
             )
 
             if response.finish_reason == "stop" or not response.tool_calls:
@@ -182,7 +184,7 @@
                 accumulated = ""
                 thinking_active = False
 
-                async for chunk in llm.stream(final_messages, temperature=profile.temperature):
+                async for chunk in llm.stream(final_messages, temperature=profile.temperature, model=profile.model):
                     if chunk.thinking:
                         if not thinking_active:
                             thinking_active = True
@@ -247,8 +249,8 @@
                 pass
         return result
 
-    def _get_backend(self, backend_key: str, model: str) -> LLMBackend:
-        return self._backends.get(backend_key, model)
+    def _get_backend(self, backend_key: str) -> LLMBackend:
+        return self._backends.get(backend_key)
 
     async def _execute_tool_calls(
         self, tool_calls: list[ToolCallRequest], tools: list[Tool]
diff --git a/navi/core/registry.py b/navi/core/registry.py
index 725bddd..9cd133d 100644
--- a/navi/core/registry.py
+++ b/navi/core/registry.py
@@ -80,7 +80,7 @@
     def register(self, key: str, backend: LLMBackend) -> None:
         self._backends[key] = backend
 
-    def get(self, key: str, model: str | None = None) -> LLMBackend:
+    def get(self, key: str) -> LLMBackend:
         backend = self._backends.get(key)
         if backend is None:
             raise KeyError(f"LLM backend '{key}' not registered")
diff --git a/navi/core/session.py b/navi/core/session.py
index 69004a0..4588e22 100644
--- a/navi/core/session.py
+++ b/navi/core/session.py
@@ -2,7 +2,7 @@
 
 import uuid
 from abc import ABC, abstractmethod
-from datetime import datetime
+from datetime import datetime, timezone
 
 from pydantic import BaseModel, Field
 
@@ -14,8 +14,8 @@
     profile_id: str
     messages: list[Message] = Field(default_factory=list)
     pinned: bool = False
-    created_at: datetime = Field(default_factory=datetime.utcnow)
-    last_active: datetime = Field(default_factory=datetime.utcnow)
+    created_at: datetime = Field(default_factory=lambda: datetime.now(timezone.utc))
+    last_active: datetime = Field(default_factory=lambda: datetime.now(timezone.utc))
 
 
 class SessionStore(ABC):
@@ -51,14 +51,14 @@
         return self._sessions.get(session_id)
 
     async def save(self, session: Session) -> None:
-        session.last_active = datetime.utcnow()
+        session.last_active = datetime.now(timezone.utc)
         self._sessions[session.id] = session
 
     async def list_all(self) -> list[Session]:
         return sorted(
             self._sessions.values(),
-            key=lambda s: (not s.pinned, s.last_active),
-            reverse=False,
+            key=lambda s: (s.pinned, s.last_active),
+            reverse=True,
         )
 
     async def delete(self, session_id: str) -> bool:
diff --git a/navi/core/sqlite_session_store.py b/navi/core/sqlite_session_store.py
index 736af6b..a28022f 100644
--- a/navi/core/sqlite_session_store.py
+++ b/navi/core/sqlite_session_store.py
@@ -2,7 +2,7 @@
 
 import json
 import sqlite3
-from datetime import datetime
+from datetime import datetime, timezone
 
 import aiosqlite
 
@@ -57,7 +57,7 @@
         return self._row_to_session(row) if row else None
 
     async def save(self, session: Session) -> None:
-        session.last_active = datetime.utcnow()
+        session.last_active = datetime.now(timezone.utc)
         messages_json = json.dumps(
             [m.model_dump(mode='json', exclude_none=True) for m in session.messages],
             ensure_ascii=False,
diff --git a/navi/llm/base.py b/navi/llm/base.py
index 4f2c765..f4d4a71 100644
--- a/navi/llm/base.py
+++ b/navi/llm/base.py
@@ -68,6 +68,7 @@
         messages: list[Message],
         tools: list[ToolSchema] | None = None,
         temperature: float = 0.7,
+        model: str | None = None,
     ) -> LLMResponse:
         """Single-shot completion. Used in the agent tool-calling loop."""
 
@@ -76,5 +77,6 @@
         self,
         messages: list[Message],
         temperature: float = 0.7,
+        model: str | None = None,
     ) -> AsyncGenerator[LLMChunk, None]:
         """Streaming text completion (no tool calling). Used for final response streaming."""
diff --git a/navi/llm/ollama.py b/navi/llm/ollama.py
index cf85016..ccefd79 100644
--- a/navi/llm/ollama.py
+++ b/navi/llm/ollama.py
@@ -30,8 +30,8 @@
     return [t.model_dump() for t in tools]
 
 
-def _base_options() -> dict:
-    opts: dict = {"temperature": 0.7, "num_ctx": settings.ollama_num_ctx}
+def _base_options(temperature: float) -> dict:
+    opts: dict = {"temperature": temperature, "num_ctx": settings.ollama_num_ctx}
     if settings.ollama_think:
         opts["think"] = True
     return opts
@@ -47,15 +47,13 @@
         messages: list[Message],
         tools: list[ToolSchema] | None = None,
         temperature: float = 0.7,
+        model: str | None = None,
     ) -> LLMResponse:
         try:
-            opts = _base_options()
-            opts["temperature"] = temperature
-
             kwargs: dict = {
-                "model": self.model,
+                "model": model or self.model,
                 "messages": _to_ollama_messages(messages),
-                "options": opts,
+                "options": _base_options(temperature),
                 "stream": False,
             }
             if tools:
@@ -89,15 +87,13 @@
         self,
         messages: list[Message],
         temperature: float = 0.7,
+        model: str | None = None,
     ) -> AsyncGenerator[LLMChunk, None]:
         try:
-            opts = _base_options()
-            opts["temperature"] = temperature
-
             async for chunk in await self._client.chat(
-                model=self.model,
+                model=model or self.model,
                 messages=_to_ollama_messages(messages),
-                options=opts,
+                options=_base_options(temperature),
                 stream=True,
             ):
                 thinking = getattr(chunk.message, "thinking", None) or None
diff --git a/navi/llm/openai_backend.py b/navi/llm/openai_backend.py
index 72dc338..0b4ba77 100644
--- a/navi/llm/openai_backend.py
+++ b/navi/llm/openai_backend.py
@@ -21,6 +21,7 @@
         messages: list[Message],
         tools: list[ToolSchema] | None = None,
         temperature: float = 0.7,
+        model: str | None = None,
     ) -> LLMResponse:
         raise NotImplementedError("OpenAI backend not yet implemented")
 
@@ -28,6 +29,7 @@
         self,
         messages: list[Message],
         temperature: float = 0.7,
+        model: str | None = None,
     ) -> AsyncGenerator[LLMChunk, None]:
         raise NotImplementedError("OpenAI backend not yet implemented")
         yield  # makes this a generator
diff --git a/navi/profiles/secretary.py b/navi/profiles/secretary.py
index d9f65b6..0611b1e 100644
--- a/navi/profiles/secretary.py
+++ b/navi/profiles/secretary.py
@@ -16,6 +16,6 @@
 
 Output style: concise, structured. When researching, include sources. Match tone and format to what was asked.""",
     enabled_tools=["web_search", "http_request", "filesystem", "code_exec", "terminal", "image_view", "reload_tools", "write_tool", "list_tools", "tool_manual"],
-    model="gemma4:e4b-it-q8_0",
+    model="gemma4:e2b-it-q8_0",
     temperature=0.7,
 )
diff --git a/navi/profiles/server_admin.py b/navi/profiles/server_admin.py
index 05bd752..aee9a39 100644
--- a/navi/profiles/server_admin.py
+++ b/navi/profiles/server_admin.py
@@ -18,6 +18,6 @@
 Workflow: gather data first (logs, status, metrics), diagnose, then act.
 Before destructive or irreversible operations, state what you're about to do and why.""",
     enabled_tools=["terminal", "filesystem", "http_request", "web_search", "ssh_exec", "image_view", "reload_tools", "write_tool", "list_tools", "tool_manual"],
-    model="gemma4:e4b-it-q8_0",
+    model="gemma4:e2b-it-q8_0",
     temperature=0.2,
 )
diff --git a/navi/profiles/smart_home.py b/navi/profiles/smart_home.py
index 4bc0770..9e4f507 100644
--- a/navi/profiles/smart_home.py
+++ b/navi/profiles/smart_home.py
@@ -18,6 +18,6 @@
 Before writing any HA config to disk, validate structure in code_exec.
 Before toggling devices or triggering automations, confirm if the action is irreversible.""",
     enabled_tools=["http_request", "filesystem", "code_exec", "terminal", "ssh_exec", "image_view", "reload_tools", "write_tool", "list_tools", "tool_manual"],
-    model="gemma4:e4b-it-q8_0",
+    model="gemma4:e2b-it-q8_0",
     temperature=0.3,
 )