Opus 4.7 to 4.6 Inflation is ~45%

A crowdsourced calculator is lighting up the feeds: community tests say Anthropic’s newer Opus 4.7 version is chewing through about 45–47% more “tokens” than 4.6 for the same asks, which means pricier chats. Tokens are the tiny pieces of text AI services bill you for—so more tokens = higher bill. The tool (not affiliated with Anthropic) shows 31 submissions averaging a 47.2% jump, with some small prompts reportedly doubling in cost.

Cue the drama. One user gasped that short messages can “2x,” while another nitpicked the title direction (“is it 4.6 to 4.7 or the other way?”) because, yes, we’re now arguing about prepositions while our wallets cry. The hottest take? A doomer forecast that companies like Anthropic and OpenAI, running big losses, will push pricier plans for tiny gains. Another commenter went full roast, saying 4.7 feels “dumb” compared to 4.6—claiming the model got “lobotomized” as safety rules tightened. And the tinfoil-hat-but-maybe-not crowd warned this is part of a bigger shift: per-token billing, lock-in via shiny tools, and a looming “monopoly/duopoly” future.

In between the panic and popcorn, a few voices add nuance: results might vary with longer answers, and the sample size is still small. Still, the memes are flying—“my AI is eating tokens like chips,” “paying more for fewer brain cells,” and “upgrade vibes, downgrade value.” Bottom line: users feel like they woke up to an unexpected AI tax, and they’re loudly asking if sticking with 4.6 is the real pro move.