@s_batzoglou
Report on my first smoke tests on Fable 5. Tested it on 5 induction problems. Fable 5 xhigh: 5/5 empty responses "", 5x128k tokens consumed and billed. Drop to Fable 5 high: 5/5 empty responses "", 5x128k tokens consumed and billed. Drop to Fable 5 medium. Report below. 1/5 nonempty responses, wrong answer. I'll do one more run, on medium (so as not to be wasting tokens) and on easier problems. (The problem set is https://t.co/gBelIZRbZI)