Anthropic releases Fable 5, the first public Mythos-class model

Anthropic released its latest model Tuesday afternoon, heralding the public’s first access to the AI company’s most powerful class of AI systems.

The company says the model, termed Fable 5, is the first publicly available product in the same family as Anthropic’s powerful Mythos models, which sent shockwaves through the cybersecurity world earlier this year for their superhuman ability to find and exploit cyber vulnerabilities.

“Fable’s capabilities exceed those of any model we’ve ever made generally available,” the company said in a blog post announcing the model’s release. “It is state-of-the-art on nearly all tested benchmarks of AI capability, showing exceptional performance in software engineering, knowledge work, vision, scientific research, and many other areas.”

According to the company, Fable 5 uses the same tier of technology as Mythos but is safe for use by the public because of safeguards and limits placed on the technology. Anthropic’s Mythos Preview model, which lacks similar safeguards, was able to find thousands of critical and severe cyber vulnerabilities, including bugs and exploits in all major operating systems and web browsers.

Many AI researchers worry that increasingly powerful AI systems could help bad actors carry out cyberattacks on banks, power grids or other critical infrastructure. Others hypothesize that increasingly intelligent AI systems could help terrorists design and deploy bioweapons.

“The same queries that are beneficial in the hands of cybersecurity professionals and biology researchers could be dangerous if available to malicious actors,” the company wrote in its blog.

To address these potential threats, Anthropic has said Fable 5 is being deployed with guardrails that block many of its responses to queries regarding potentially dangerous topics. For those user prompts, Anthropic will steer answers to an earlier, less-powerful model called Opus 4.8 — which was the highest-performing publicly available model until Tuesday.

“Fable 5’s capabilities in areas like cybersecurity, biology and chemistry are advanced enough that we’re taking a deliberately conservative approach for these topics at launch,” Anthropic wrote in reply to a question from NBC News. “To enable general availability of other Mythos level capabilities, we’ve decided to deploy safeguards that err on the side of caution, applying broad restrictions to these topics for now.”

Anthropic said the two-pronged approach, diverting sensitive questions from Fable 5 to the older Opus 4.8, would allow users to still obtain helpful answers to questions when Fable 5’s capabilities could prove too dangerous.

Anthropic said it will make Fable 5 available to all users on its Pro, Max, Team and seat-based Enterprise plans at no additional cost until June 22. After that, users will only be able to access the model by buying and utilizing extra computing credits.

Anthropic said it aimed to incorporate Fable 5 into usual subscription plans “as quickly as we can.” Anthropic has recently struggled to keep up with soaring demand for its AI systems, and Fable 5 is likely to add additional stress to its limited computational resources.

Anthropic also announced that the trusted partners who had previously been able to access Mythos Preview will now be able to access an upgraded Mythos model, called Mythos 5.

The company had made Mythos Preview available to over 150 organizations around the world to help financial institutions, software companies and healthcare networks shore up weaknesses in their cyber defenses prior to a wider public release of a Mythos-class model.

Anthropic said Mythos 5’s capabilities shattered existing performance records across a range of other domains, including drug design and molecular biology. Anthropic said Mythos 5 is the company’s “first model to consistently produce novel, compelling scientific hypotheses.”

Tuesday’s model release comes one week after President Donald Trump signed a new executive order that aims to establish a new, voluntary mechanism for AI companies to share their systems with the government for safety testing before they are publicly released and to shore up the government’s own cyber defenses.

Source link