Anthropic Opens the Gates to Its Most Powerful AI Model Yet - With Strings Attached

After months of speculation and carefully controlled access, Anthropic has finally released a version of its next-generation AI that businesses and developers can actually use. Meet Claude Fable 5, the first commercially available model from the company's Mythos family.

Anthropic positioned the original Mythos Preview as a cybersecurity powerhouse so potent it could theoretically break the internet if misused. That's why the company kept it locked down. Now, Fable 5 arrives as a safety-tuned alternative, designed for software engineering, knowledge work, and image understanding - including a notable first: it beat Pokemon FireRed using only vision inputs, something earlier models couldn't manage.

But accessing this power comes with trade-offs. Fable 5 requires significantly more compute than Anthropic's Opus line, so the company is limiting availability. Subscribers on Pro, Max, Team, or Enterprise plans can use it now through June 22. After that, it disappears from standard plans and moves to a pay-as-you-go credit system, with API pricing at $10 per million input tokens and $50 per million output tokens.

On safety, Anthropic claims Fable 5 rejected every single harmful request related to cyberattack planning, exploit development, or defense evasion - even when tested with 30 different jailbreak techniques. To prevent potential bioterror risks, the model won't answer most biology or chemistry questions, routing those to the older Opus 4.8 instead. The company is also working with the U.S. government on eventual plans for the full Mythos model, and business users must now agree to a 30-day data retention policy for misuse monitoring, not model training.

For now, Fable 5 is a glimpse of what's possible - if you're willing to pay for the privilege and work within tight guardrails.

Source: CNET