Skip to content

Comment on Codex (GPT-5.5, Plus plan) – rate-limit cost per token jumped 10x+ since June 16parent

Comments

yes and no. Ai models become more efficient (I've seen a comparison somewhere). So even if the prices stay the same, usage could go down. Also, in the uber example, there's no free alternative, unlike ai models. I'm yet to try qwen and some others though, so can't vouch for them.

But that doesn’t seem like it would be them undercutting each other, per se. And to me it looks like most people are saying the number of tokens used per task just keeps going up, and on top of that, monthly plan pricing seems to be going away. I think we hit the summit of peak discount a few months ago, now they’re being forced to be more financially independent approaching their IPOs, and I just don’t see any factor feasibly forcing costs down after the subsidies go away. SpaceX had the rest of SpaceX to take the pressure off XAI being a money furnace, but what do OpenAI and Anthropic have?

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.