Anthropic is rolling out a public version of its Mythos AI model through Claude Fable 5, while blocking use in risky areas such as cybersecurity. The company said Tuesday that Claude Fable 5 is its most powerful model made for wider use, with stronger performance in software engineering and analytics.
# Anthropic Rolls Public Mythos Model With Cybersecurity Limits
Anthropic is rolling out a public version of its Mythos AI model through Claude Fable 5, while blocking use in risky areas such as cybersecurity. The anthropic rolls public release gives wider users access to what the company calls its most powerful generally available model, while keeping the higher-risk capabilities from the earlier Mythos preview behind restrictions.
The company said Tuesday that Claude Fable 5 is designed for broader use and highlighted its performance in software engineering and analytics. The release follows a more limited Mythos rollout earlier this year that drew attention because the model had been able to find software flaws.
What Happened
Anthropic is bringing Mythos-class capabilities to a wider audience through Claude Fable 5, but the public model will not allow risky cybersecurity use. That means users asking it to help find vulnerabilities in a package or code would be refused, according to Dianne Penn, Anthropic's head of product management, research and labs, who spoke to Reuters.
Penn said that in such a case, Fable 5 would refuse the request and fall back to Opus 4.8 for a response. Anthropic said it had carried out extensive testing to ensure that users could not manipulate the model into bypassing guidelines or performing restricted actions.
The move gives Anthropic a way to broaden access without fully opening the security-sensitive capabilities that made the earlier Mythos preview notable. It also separates the public Claude Fable 5 release from Claude Mythos 5, which remains tied to users who already had preview access.
Anthropic Rolls Public Mythos With Guardrails
The new model is positioned as the most powerful Anthropic has made available for wider use. The company is promoting Claude Fable 5 for software engineering and analytics work, two areas where advanced reasoning models are increasingly used to inspect code, process technical material and support complex workflows.
But Anthropic is drawing a line around cybersecurity. The earlier Mythos preview was limited after the company said in April that Mythos had uncovered thousands of software vulnerabilities. Since then, access has been restricted to about 200 organizations, including the US government under the Glasswing program.
For the public release, Anthropic says the guardrails are meant to prevent risky use while still giving general users stronger model performance. The company also said users with access to the preview version of Claude Mythos, which did not have the same guardrails, would be able to upgrade to Claude Mythos 5.
The Details
Pricing for both Claude Fable 5 and Claude Mythos 5 is $10 per million input tokens and $50 per million output tokens, according to the company. Anthropic said Fable 5 will be a more expensive model, but Penn said early customer feedback suggests it can complete tasks with lower token usage, reducing the overall cost per task.
That distinction matters because model pricing is not only about the listed token rate. A model that uses fewer tokens to complete a task can cost less in practice even if each token is priced higher. Anthropic's claim is based on early customer feedback cited by Penn.
The company also said it plans to expand access over time through a more systematic trusted-access programme. That signals a continued split between general availability and controlled access for capabilities Anthropic considers sensitive.
Context and Background
Mythos drew global attention earlier this year because of its reported ability to find software vulnerabilities. Anthropic responded by limiting access to a small group of organizations rather than making the full capability broadly available from the start.
The new release keeps that caution in place. Claude Fable 5 gives wider users access to a more capable general model, but the public version is not framed as an unrestricted cybersecurity tool. Instead, Anthropic is using refusal behavior and fallback responses to keep some requests outside the model's permitted use.
The company is also operating in a market where leading AI startups are racing to scale usage, raise valuations and prepare for possible public listings. Dawn's Reuters report noted that broader access could help Anthropic extend momentum that has lifted its valuation above rival OpenAI.
Why It Matters
The release shows how leading AI labs are trying to balance broader access to stronger models with controls around high-risk use cases. Anthropic is not simply withholding Mythos-class capability from the public, but it is limiting the parts most closely tied to cybersecurity risk.
For customers, the practical question is whether Claude Fable 5 can deliver better software engineering and analytics results while respecting those restrictions. For the broader AI market, the launch is another example of model makers using product tiers and trusted-access programmes to separate general productivity use from sensitive technical capabilities.
Key Takeaways
- Anthropic is releasing Claude Fable 5 as a public version of Mythos-class capability.
- The wider-use model blocks risky cybersecurity requests, including vulnerability-finding prompts.
- Mythos access had previously been limited to about 200 organizations, including the US government under Glasswing.
- Claude Mythos preview users can upgrade to Claude Mythos 5.
- Anthropic says both models are priced at $10 per million input tokens and $50 per million output tokens.
*Source: [Dawn](https://www.dawn.com/news/2006596/anthropic-rolls-out-public-version-of-mythos-without-cybersecurity-capability)*
Frequently Asked Questions
What is Claude Fable 5?
Claude Fable 5 is Anthropic's new wider-use model that brings public access to Mythos-class capabilities with restrictions on risky cybersecurity activity.
Does Claude Fable 5 include cybersecurity capability?
No. Anthropic says the public version blocks risky cybersecurity uses, including requests to find vulnerabilities in software packages or code.
Who had access to Mythos before the public release?
Anthropic had limited access to about 200 organizations, including the US government under the Glasswing program.
How much does Claude Fable 5 cost?
Anthropic said pricing for both Claude Fable 5 and Claude Mythos 5 is $10 per million input tokens and $50 per million output tokens.
What happens to Claude Mythos preview users?
Users who had access to the preview version of Claude Mythos can upgrade to Claude Mythos 5, according to the company.
The Bottom Line
Claude Fable 5 broadens access to Anthropic's Mythos-class AI while keeping cybersecurity-sensitive capabilities restricted. The release gives Anthropic a wider-use model for software engineering and analytics, alongside a controlled path for users who already had Mythos preview access.
Continue reading related coverage in News or browse all stories on the articles page.