@s_batzoglou
My first test of Fable 5 on some hard reasoning tasks (induction) didn't go so well. 5/5 responses are "" after billing 128k tokens. Glad I didn't run the full 375-problem benchmark. How many tokens does it need to return anything other than ""? https://t.co/paZfTvZGZW https://t.co/9MONVXax20