[JellyYelly]

They say its mythos like, without actually comparing it to Mythos (fair enough, it's not public) but the bar for a model to be mythos-like has to be that you can produce as many novel and high severity security vulns outlined in the Mythos redteam blog. I haven't seen any other lab produce a report like that yet. The proof is in the pudding.

[cassianoleal]

> The proof is in the pudding.

Funny you say that, when the Mythos team have produced no proof either.

[subscribed]

Not sure if the reports like this count? https://www.theregister.com/2026/04/22/mozilla_firefox_mytho...

I don't have strong opinion on that.

[maplethorpe]

I believe they've stated that it would be too dangerous to release.

[satvikpendem]

Just like OpenAI said GPT 2 was too dangerous to release?

There was just an article on this phenomenon today: https://news.ycombinator.com/item?id=47890235

[maplethorpe]

They released a system card talking about how powerful it was. I don't think OpenAI did that with GPT 2.

[satvikpendem]

I mean, that's just part of the marketing too. OpenAI would've absolutely added a system card, they just weren't invented back in the GPT 2 era.

[Razele]

too uneconomical to run*

[halJordan]

It benches very similarly on the cyber benches Anthropic put out. That meets the bar.