Rumors that OpenAI recently finished new >10T parameter training run

5 pointsposted 11 hours ago
by andy12_

1 Comments

sscaryterry

10 hours ago

> All of this has Anthropic concerned. OpenAI believe Anthropic have no good response to Astra prepared for a public launch, and my understanding is that they're correct - mostly as a result of compute constraints, Anthropic are bracing for much of the rest of the year to go OpenAI's way, but think they'll be back on top early next year.

Well, Anthropic is simply losing because they seem to hate their customers, whereas OpenAI seem love them right now.

Getting awesome results with Sol 5.6 at high, consistently. Not only that, frequent resets.

On the other hand, I tried Opus 5 for 3 days, after which I wanted to give up this life and farm pumpkins. :)