Anthropic just launched Claude Fable 5, its most powerful AI model open to the public. This marks a big shift. For the first time, everyday developers and businesses can tap into Mythos-level AI technology.

Fable 5 excels at tough tasks like coding, research, and visual understanding. It can handle long, complex projects better than earlier Claude models. It keeps context over millions of words, which helps it stay focused and accurate.

One impressive example: Stripe used Fable 5 to migrate 50 million lines of Ruby code in just one day. Normally, that would take months for a whole team. This shows how the model can speed up software development.

Safety First with New Guardrails

Anthropic built strong safety controls into Fable 5. The model blocks sensitive or risky questions in areas like cybersecurity, biology, and chemistry. When these come up, it switches those queries to a safer, older model called Claude Opus 4.8.

This fallback happens in fewer than 5% of user sessions, so most people get direct answers from Fable 5. Anthropic tested the model with over 1,000 hours of hacking attempts and bug bounties. No universal jailbreaks were found, but they stay cautious.

To catch new threats, Anthropic now keeps user data for 30 days. This helps them spot and block complex attacks or harmful queries. The data is not used to train the model but only for security. This policy is new for top-tier AI access.

Introducing Claude Mythos 5 for Trusted Partners

Alongside Fable 5, Anthropic launched Claude Mythos 5. It is the same core AI but with fewer restrictions. Mythos 5 is only available to select cybersecurity experts and biology researchers through a trusted access program called Project Glasswing.

Mythos 5 has shown faster protein design and new biology insights during internal tests. It even analyzed single-cell data from 138 animal species, outperforming recent scientific studies. This highlights its potential in life sciences research.

The restricted access to Mythos 5 aims to balance power with control. Anthropic wants to support high-security applications and advanced research while limiting risks of misuse.

Access and Pricing

Fable 5 is available now through Anthropic’s Claude API and subscription plans like Pro, Max, Team, and Enterprise. Users can try it at no extra cost until June 22. After that, it requires usage credits until Anthropic reintegrates it as a standard subscription feature.

The pricing is $10 per million input tokens and $50 per million output tokens. This is less than half the cost of the earlier Mythos Preview but still twice as expensive as Opus 4.8. Some companies might find that cost a challenge.

Despite the price, companies like Rakuten see great value. They say Fable 5’s ability to evaluate and validate its own work enables truly autonomous operations. This extra thinking power saves time and effort.

Fable 5 also shines in vision tasks. It can read scientific charts, understand screenshots, and even rebuild web apps from images. It even completed a Pokémon game using only screenshots, without maps or extra info.

With this launch, Anthropic joins the race to bring frontier AI to more users. They stress the importance of careful safety measures as AI technology advances rapidly. The company also plans a possible IPO soon, aiming to grow while keeping control.

Overall, Claude Fable 5 offers developers a powerful yet cautious step into next-generation AI. It combines cutting-edge capabilities with strong guardrails to keep users and society safe.