[louiereederson]

For a 56.7 score on the Artificial Intelligence Index, GPT 5.5 used 22m output tokens. For a score of 57, Opus 4.7 used 111m output tokens.

The efficiency gap is enormous. Maybe it's the difference between GB200 NVL72 and an Amazon Tranium chip?

[swyx]

why would chip affect token quantity. this is all models.

[louiereederson]

Chip costs strongly impact the economics of model serving.

It is entirely plausible to me that Opus 4.7 is designed to consume more tokens in order to artificially reduce the API cost/token, thereby obscuring the true operating cost of the model.

I agree though, I chose poor phrasing originally. Better to say that GB200 vs Tranium could contribute to the efficiency differential.

[itemize123]

probably the wrong take - they are arm racing to a better model. it's not enshittification era for models just yet

[fiatpandas]

Models are still in arms race mode, but harnesses and subscription strategy are tiptoeing into their enshittification era.

[dist-epoch]

If it's a new pretrain, the token embeddings could be wider - you can pack more info into a token making it's way through the system.

Like Chinese versus English - you need fewer Chinese characters to say something than if you write that in English.

So this model internally could be thinking in much more expressive embeddings.

[AtNightWeCode]

You need to compare total cost. Token count is irrelevant.

[karmasimida]

Chips doesn’t impact output quality in this magnitude

[ChrisGreenHeur]

True, but the qualifying the power played a large part. Most likely nuclear power for this high quality token efficiency.