Braintrust which is a really solid eval tool/platform just compared it to Opus 4...

		scritty-dev 2 days ago \| parent \| context \| favorite \| on: GLM 5.2 and the coming AI margin collapse 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
		help