Claude just released a Mythos-level model, but you only have 10 days to try it

Summary

• Anthropic has released Fable 5 — a Mythos-level Claude model available for a limited time.
• Tuned for cybersecurity and biology and features robust safeguards to drop it to cheaper models if triggered.
• Subscribers can use Fable 5 until June 22; expect high cost and token burn.

If you've been itching to get your hands on Claude's Mythos model — the company's top-tier model designed for deeper coding and even spotting bugs and cybersecurity issues in other apps and systems — then now might be your only chance, at least for a little while. That's because Anthropic has hot-dropped a new model for Claude subscribers to test out called Fable 5, and it's only out for a limited time.

A taste of what Claude can really do

It's Mythos-level, supposedly

According to Anthropic's announcement post about Fable 5, the new model appears to be at least on some level similar to Mythos, as it is actually labeled next to Mythos 5 on a performance graph. The company says it is built to tackle your longest-running and most complex usages. Additionally, based on the performance graph seen in the announcement, it even performs better than the Mythos Preview in some cases.

Now, none of this is to downplay how impressive Claude can be, especially if you utilize it correctly. For instance, using Claude Cowork to take over repetitive tasks can be game-changing.

If you haven't been paying attention to the conversation surrounding Claude Mythos, the model is supposedly so dangerously good at what it does, that Anthropic is afraid to release it to the general public. Instead, the company has been testing it behind closed doors, as part of a larger cybersecurity push to bring more AI into that particular sector of things. That, itself, is terrifying enough. However, Fable 5 appears to be a more consumer-focused version of the Mythos model. Though, exactly how much has changed, or how much Anthropic has limited it, is hard to say without a direct comparison.

However, the model is clearly more geared towards cybersecurity and biology industries, as Anthropic even notes in the announcement that there are "robust safeguards for cybersecurity and biology" that will kick the AI model to a less expensive model should any of the safeguards be hit. This should help cut down on usage, which is twice the cost of Claude Opus, according to a note on the Claude website and app.

A limited time to try it

After June 22, you'll need to dish out extra credits for access

Which brings us to the biggest crux of this announcement. Despite the supposed advancements that Anthropic is flaunting on its website for Fable 5, access to the new model is going to be limited. Subscribers to Claude can try it out right now, but not only is it expensive (as noted above), but it will stop being available on June 22. That means you have just over 10 days to toy around with it before Anthropic takes it off the list and locks it behind additional credit purchases.

We also don't really know how well it will stand up against Mythos in day-to-day usage, as nobody in the general public has had a chance to really toy with Mythos yet, thanks to its extremely limited access. And that doesn't seem to be changing, as Mythos 5 will continue to only be available to Glasswing members (which is the cybersecurity alliance that Anthropic set up with several other major tech companies). But it is pretty powerful, as Anthropic has already shown off that it can beat Pokémon with just visual input.

If you're Claude Pro subscriber, then this is definitely one of those Claude features you need to take advantage of, especially if you like messing around with the latest and greatest AI models, then Claude has a new one to try. Just don't be surprised if you run out of tokens within a couple of prompts.