[alach11]

I ran an internal (oil and gas focused) benchmark yesterday and found Opus 4.7 was 50% cheaper than Opus 4.6, driven by significantly fewer output tokens for reasoning. It also scored 80% (vs. 60%).

[stingraycharles]

That’s just adaptive reasoning, not related to the increased tokenizer costs.

[simianwords]

Why would I as a user be concerned about one over the other?

[stingraycharles]

Because it teaches you cause and effect in terms of costs and quality.

Unless you want to keep complaining about the model being nerfed.