@weswinder Yeah, very weird. They're not advertising any tps improvements either so this just feels like a flop out the gate. Not something you want to see from anthropic. There was talks about them being compute constrained, so maybe this is just more efficient... but the tokens. idk weird
Sonnet 5 at the top of this @ArtificialAnlys cost chart has got to be a big wtf moment for most people. More than opus, more than double GPT 5.5.
They didn't raise the price either. This is just an insane amount of tokens used. Yes it's one benchmark.... but talk about yikes.
@SenSanders What is your stance on the idea that if we don’t do it first, others will and will gain a massive foot forward in this massively important industry?
@Prathkum I’ve been mentoring for seniors in their capstone project and I’ll tell you it’s the same. Except, they’re being told to use AI as much as they can for their project now. Seems strange to me.
@yacineMTB I would love an explanation on how this works across multiple providers for the model. Is this directly from anthropic and if so couldn't you just use another provider? If not, why?
@thdxr We do both. The instructions have become a bunch of guidelines/rules as well. We still struggle with consistency in longer running tasks though.
@levelsio Aren’t their multiple providers for the top models though? You can get opus 4.6 through Amazon Bedrock for example. Has anyone tested if there’s differences between providers?