Braintrust which is a really solid eval tool/platform just compared it to Opus 4.8 to see if it could preserve exact long context retrieval under prod serving c

Braintrust which is a really solid eval tool/platform just compared it to Opus 4.8 to see if it could preserve exact long context retrieval under prod serving constraints and it did really well. I think 6-12 months before OSS has Fable-esque models