Anthropic is gradually widening the gap with its competitors with its highly intelligent models. Photo: ZDNET .
Early morning on June 10th ( Hanoi time), Anthropic launched Claude Fable 5, the first widely released version of its Mythos layer model. The company stated that Fable 5 excels in software engineering, knowledge work, and computer vision, but has significant security limitations. In high-risk fields such as cybersecurity, biology, and chemistry, the model will automatically block responses and redirect them to Opus 4.8 for processing.
Alongside Fable 5 is Mythos 5, sharing the same core technology but with certain security barriers removed, intended for a small group of security companies.
Launched as a preview in April, Mythos was initially limited to a small group of partners due to cybersecurity concerns. Last week, Anthropic expanded access to hundreds of organizations in 15 countries, continuing its focus on critical infrastructure managers.
Fable 5's benchmark scores compared to the best models currently available. Image: Anthropic.
Currently, a version of that technology is available to all customers via the Claude API and Enterprise plans that charge per usage. Access on monthly subscription plans will be rolled out in phases. From now until June 22nd, Fable 5 will be integrated free of charge into the Pro, Max, and Team plans. On June 23rd, Anthropic will remove Fable 5 from these plans, requiring users to purchase additional credits to continue.
The launch of Fable comes as Anthropic prepares for an IPO, coinciding with OpenAI and Elon Musk's SpaceX. This move follows Anthropic's call for major AI labs worldwide to establish a "brake" for the development of frontier AI. The firm warns that systems are advancing too quickly, to the point where they may soon achieve self-upgrading capabilities.
In third-party tests, analytics firm Hex shared in a statement that Fable was the first model to achieve 90% of their core analytics benchmark for complex, long-running tasks.
"On the most difficult questions, it demonstrates keen judgment and attention to subtle nuances," Hex commented.
The vibe-coding platform Base44 noted that Fable excelled at "generating entire applications with just one statement" and had excellent tool-calling capabilities. Meanwhile, the AI solution Genspark stated that Fable beat every other model in their reviews and performed significantly better in tasks such as user interface (UI) design and game programming.
The price for both Fable 5 and Mythos 5 is $10 per million tokens invested and $50 per million tokens sold, double the price of Opus 4.8. That price alone could be a barrier to widespread adoption.
In the benchmarks published by Anthropic, Fable 5/Mythos 5 outperformed the rest of the market across all metrics. The results are even more convincing when compared to Opus 4.8, a model highly regarded for its intelligence, performance, and especially its superior programming capabilities compared to comparable options from OpenAI or Google.