@tjrobertson52

6. I wanna talk about what we know so far and help you decide if you need to start thinking about switching over from Claude or Gemini. 6 models: Soul, Terra, and Luna.

None of them are publicly available yet, so we just have to go off of what OpenAI has released about each of the models. 5 at half the cost. And of course Luna will be even faster and cheaper, but not as capable.

This is one area where I think OpenAI is legitimately winning against Anthropic. Anthropic continues to put out smarter and more capable models, but always at the same or higher cost than their previous models. Everyone who's serious about AI right now wants to use Anthropic's models, but the elephant in the room is that it's getting harder and harder to afford them.

OpenAI is addressing this head on. 8. 6 Terra is just as good at that at half the cost, a lot of people are gonna be switching over.

But of course, OpenAI is not comfortable being just the cheaper version of Claude. 6 Soul as comparable to Mythos. However, so far I'm not convinced.

There's two problems with OpenAI's announcement that make it unconvincing. One is, so far they've only shared a handful of benchmarks. As we all know, the model makers always pick the benchmarks where their model does the best.

And two, on some of the benchmarks, they don't even show where Mythos is. And when they do show where Mythos is, it's Mythos preview, which is a model that's over two months old now. 5 is alrea

And when they do show where Mythos is, it's Mythos preview, which is a model that's over two months old now. 6 Soul is going to be a significant jump up. But if it were truly comparable to Fable or Methods 5, I think it would be much easier for OpenAI to demonstrate it.

In their announcement, opening, I also spent considerable time talking about how they're cooperating with the United States government to mitigate any cybersecurity or biological weapon risks. Seems like this might become the new normal, at least until we find a better way of verifying the safety of these powerful models. But also it just felt like OpenAI was trying to prove that Anthropic's not the only one with a big scary model.

For now, though, it really feels like Anthropic has a significant lead on the most capable frontier models. Of course, anything could happen, but I don't see anyone catching up to them anytime this year. OpenAI might be winning on cost, but keep in mind if you're paying for a subscription to either ChatGPT or Claude, you're getting a subsidized rate.

You're only paying between 4 and 12% of the actual usage cost. If the work you're doing can be done inside of one of these subscriptions, I still think Claude is currently the best choice. And if you're doing big, important, ambitious projects and you need the smartest model, then you need to be using Claude right now.

And it looks like we'll be getting Fable 5 back any day now.