Use Nemotron 3 Nano Omni reasoning model

mp1/pluto/modes.py

@@ -82,7 +82,7 @@ def _build_registry() -> dict[str, ModeConfig]:
             ),
             "MODE_REASONING": ModeConfig(
                 mode_name="MODE_REASONING",
-                model_id="nvidia/llama-3.3-nemotron-super-49b-v1",
+                model_id="nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
                 temperature=0.3,
                 max_tokens=4096,
                 compute_profile="high-reasoning",
@@ -107,7 +107,7 @@ def _build_registry() -> dict[str, ModeConfig]:
             # Keep MODE_GEMINI name for backward compat — maps to Super
             "MODE_GEMINI": ModeConfig(
                 mode_name="MODE_GEMINI",
-                model_id="nvidia/llama-3.3-nemotron-super-49b-v1",
+                model_id="nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
                 temperature=0.0,
                 max_tokens=4096,
                 compute_profile="high-throughput",