[cheeseblubber]

Via api you can turn off websearch internally. We provided all the models with their own custom tools that only provided data up to the date of the backtest.

[mlmonkey]

But Grok is internally training on Tweets etc. continuously.