Anthropic has made Claude Fable 5 available to the public through the Claude desktop app, giving paid subscribers their first look at what the company calls a Mythos-class AI model. The release marks a significant step in the company's public rollout strategy, but access comes with conditions that users should understand before diving in.
What Is Claude Fable 5?
<a href="https://www.analyticsinsight.net/news/anthropic-launches-claude-fable-5-with-advanced-safety-controls" target="_blank" rel="none noopener">Fable 5</a> is described by Anthropic as a Mythos-class model, a tier that had previously been withheld from general availability due to concerns around its capabilities. Anthropic says the new model shows major improvements in coding, spatial reasoning, and cybersecurity, with the biggest performance jumps coming in the cybersecurity category.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
Anthropic's press release states that the model has the potential to do profound good for the world, but that framing comes alongside a direct acknowledgment of risk. The company warns that Fable 5's extreme cybersecurity capabilities could be misused to cause serious damage, which is why the public version ships with conservative safety guardrails in place.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
Safety Guardrails and the Mythos 5 Split
The Fable 5 that most users will interact with is a deliberately restrained version of the underlying model. Anthropic acknowledges that these safeguards may sometimes catch harmless requests, though it estimates this happens in fewer than 5 percent of sessions. That is a notable admission, and users doing technical or research-oriented work may occasionally find the model more restrictive than expected.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
For organizations working in cybersecurity, Anthropic is taking a different approach. A separate release called Claude Mythos 5, based on the same underlying model but with safeguards lifted in some areas, is being made available to a small group of cyberdefenders and infrastructure providers. This version will not be accessible to the general public.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
Who Can Access Fable 5 and for How Long
Paid <a href="https://www.analyticsinsight.net/artificial-intelligence/claude-code-vs-codex-how-to-combine-both-ai-tools-effectively" target="_blank" rel="none noopener">Claude subscribers</a>, including those on Pro, Max, Team, and seat-based Enterprise plans, can use Fable 5 right now through the Claude desktop app. However, that window is short. Anthropic will remove Fable 5 from standard subscription plans on June 23, after which users will need paid usage credits to continue accessing it. These credits are prepaid and billed at the same rate as API usage.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
Anthropic says it aims to restore Fable 5 as a standard part of subscription plans as soon as possible, but only when sufficient capacity allows. There is no confirmed date for that to happen.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
What to Expect in Practice
Users who try Fable 5 before the June 23 cutoff should note that it consumes usage credits significantly faster than earlier models. Anthropic's next most powerful offering, Claude Opus 4.8, will remain available as the fallback option for subscribers once Fable 5 is pulled from standard plans.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
Early reports suggest the rollout has not been entirely smooth. At least one writer attempting to use Fable 5 with Claude Code was quickly met with a vague error message indicating an issue with the model.<a href="https://www.pcworld.com/article/3161065/claude-too-dangerous-ai-model-is-finally-public-but-theres-a-catch.html"> </a>
The Bigger Picture
The release of Fable 5 reflects a broader tension Anthropic has been managing publicly: how to release increasingly capable models while being transparent about their risks. Offering a scaled-back public version while reserving a less restricted build for trusted security professionals is one answer to that question, though it raises its own debates about access and accountability. For now, users have until June 23 to test one of the most capable AI models Anthropic has ever shipped.

Anthropic launches Claude Fable 5, its first public Mythos-class AI model, with advanced capabilities and safety controls.

Anthropic’s Claude Fable 5: What is it, Safety Guardrails & How to Use It?

Other Articles

Mid-Day Fast

Home / Buzzfeed / Article / Anthropic’s Claude Fable 5: What is it, Safety Guardrails & How to Use It?

Anthropic’s Claude Fable 5: What is it, Safety Guardrails & How to Use It?

Other Articles

Mid-Day Fast