diff --git a/server/routes.go b/server/routes.go
index cdd3caf58..5c52dbfef 100644
--- a/server/routes.go
+++ b/server/routes.go
@@ -112,6 +112,10 @@ func load(ctx context.Context, workDir string, model *Model, reqOpts map[string]
 		loaded.Options = &opts
 	}
 
+	// update options for the loaded llm
+	// TODO(mxyng): this isn't thread safe, but it should be fine for now
+	loaded.runner.SetOptions(opts)
+
 	loaded.expireAt = time.Now().Add(sessionDuration)
 
 	if loaded.expireTimer == nil {