Use GPT-4.1 as the default model

It seems to perform the best in accelerant, as compared to both 4o and
the reasoning models. It also is the default for the OpenAI Agents SDK,
probably for similar reasons.

Author: camelid

accelerant_server.py

@@ -18,7 +18,7 @@ def route_optimize() -> str:
     filename = request.args.get("filename")
     lineno = request.args.get("line", type=int)
     perf_data_path = request.args.get("perfDataPath", type=Path)
-    model_id = request.args.get("modelId", "o3-mini")
+    model_id = request.args.get("modelId", "gpt-4.1")
 
     if (filename is None or lineno is None) and perf_data_path is None:
         raise Exception(