Anthropic, the company behind Claude, has unveiled new versions of its two flagship models, Fable 5.1 and Mythos 5.1, as it looks to regain the lead in the artificial intelligence development race. In a statement, the company led by Dario Amodei struck an ambitious tone about both models, explaining that “their research capabilities offer an early glimpse of how AI models will contribute to scientific progress.”
In reality, Fable 5.1 is the one that matters most to us, since Mythos will remain reserved for a very specific audience: tech giants and international organizations that use it in the field of cybersecurity. In fact, Anthropic clarifies that Claude Mythos 5.1 is identical to Fable 5.1, but offers more permissive safety measures for verified individuals and organizations.
So, let us break down the key takeaways from the launch of Fable 5.1.
When it comes to its new flagship model, Anthropic is placing particular emphasis on its price.
Although Anthropic has decided to keep it the same as Fable 5 ($10 per million input tokens and $50 per million output tokens), it estimates that, in terms of performance, Fable 5.1 will be 25% cheaper than its previous version “for typical workloads, wherever usage is billed by token.” That is because the new model reduces the price of cache reads (where the model reads inputs that have already been processed and stored).
It also says that in highly agent-dependent tasks, savings could reach as much as 45%. “Fable 5.1 achieves results similar to or better than Fable 5’s at a much lower cost,” the company states.
But there is some fine print: this does not mean that Fable 5.1 will always be cheaper.
One of the leading model analysis and comparison platforms, Artificial Analysis, has published a brief study explaining that Fable 5.1 uses far more output tokens when operating at full capacity. The platform estimates that it generates approximately 1.7 times more output tokens than Fable 5. And output tokens are precisely the expensive ones: $50 per million.
In fact, Artificial Analysis estimates that, when operating at its highest effort level, the new model costs around $3.76 per task on its Intelligence Index, compared with $3.14 for Fable 5, 20% more.
As for its performance, Anthropic has shared benchmark results highlighting Fable 5.1’s stronger capabilities. In fact, judging by the following table, Fable 5.1 doubles Fable 5’s performance in scientific research, a major milestone in the evolution of this type of model. Of course, this is a table provided by Anthropic itself, in which it limits the comparison to Fable 5.1, its predecessor, Opus 5, and GPT 5.6 Sol, the only competing model included.
And in this case, Artificial Analysis comparisons point in the same direction: in its rankings, Fable 5.1 already stands as a leader across many of the classifications published.
Source: Anthropic
Anthropic also highlights that the new version of Fable has improved in areas such as security and privacy. For example, Fable 5.1 uses a new system called Enterprise Frontier Safeguards (EFS), designed to offer privacy to customers by storing data in a cloud infrastructure fully controlled by the customer, not by Anthropic. In any case, EFS will not begin rolling out until “later this fall.” Until then, customers will be able to use Fable 5.1 with zero data retention.
Starting today, Fable 5.1 is already available on platforms such as AWS, Google Cloud, and Microsoft Azure, and developers can also access it through the Claude API.
Your email address will not be published. Required fields are marked *
Δ