Thanks for this! Will test this model out because we do a lot of in between steps to get around the output token limits.

Super nice if it worked for our use case to simply get full output.