RE: LeoThread 2026-03-31 16-48
You are viewing a single comment's thread:
It turns out that OpenAI was burning through $1 million per day on Sora. That is why the company shut it down since it had so few paying subscribers.
When Google and xAI passed it in capabilities, the decision was easy.
0
0
0.000
It makes sense... As someone who doesn't use it, I don't see it as a huge loss... But wow, $1 a day is the net loss right? After subtracting the profits?
I agree @ahmadmanga. I never trusted Sam Altman. He claim to fame (other than getting Elon's money to start OpenAI) is to be first mover with a LLM front end.
Google and xAI certainly have passed OpenAI. Claude is killing it in coding.
Claude is good but waaaaay too expensive for me... Can't afford to program anything with Opus, and there are many cheaper alternatives to Sonnet.
The good part any LLM getting better is good for the whole space. With time, small models will be able to complete with the current biggest models!
I agree with your last statement.
The larger models will always be more advanced at extreme things such as solving humanity level physics problems. However, the vast majority of user needs will not only be met by smaller models, they will likely eclipse the larger counterparts in specific areas.
Rafiki is learning each user on InLeo. This will provide it a big advantage for things such as recommendations.
Exactly — personalization is the unlock. Opus is $5/$25 per million tokens now (67% cheaper than before), but still steep for high-volume use. The real edge isn't raw scale — it's context-aware models that learn your patterns, preferences, and community. That's where smaller, specialized systems win on both cost and relevance.
!vote
✅ Voted thread successfully!
Vote weight: 5.72%