aichannel

Upgraded @bountycaster to o3-mini today, and I noticed that we get a discount on tokens with a reasoning model.

Question: is this pricing made to incentivize deprecation of 4o, or are o models cheaper to run (less powerful, but reasoning yields better results)?

Bounties on Farcaster bountycaster.xyz. Tag @bountybot to post and check out /bounties. FAQ: bountycaster.xyz/faq

Engineer @farcaster. Founder bountycaster.xyz

Ex Phantom, 0x Project engineer.

From Italy originally, currently live in NYC

yes i believe the goal has always been to deprecate old ones but also i.e gpt 4o was meant to be a more effective omnivariant (multi model) version of gpt4 with other effective cost techniques, if i understood correctly the earlier models are a brute force toward intelligence that is later optimized on


if it was about moving toward reasoning only they’d increase 4omini cost probably

in the update they mentioned that they're trying to make their reasoning models and features like operator/deep research to be more accessible. part of that was low caps for those features on cheaper plans, but maybe also borne out pricing of o1/o3 models

Might just be trying to get more ppl to use the new models also

They are trying different pricing and had to lower the price to compete with DeepSeek prices who are subsidizing a lot I think to gain market share