Stay updated with ZDNET: Add us as a preferred source on Google.
ZDNET’s main points
- Anthropic is rolling out Claude Fable 5 to all users.
- Fable 5 delivers Mythos-level performance with built-in safety measures.
- It costs roughly double the price of Claude Opus 4.8.
Anthropic has unveiled a toned-down version of its legendary (and tightly controlled) Mythos large language model. Named Claude Fable 5, the company calls the new AI “a Mythos-class model engineered for safe, everyday use.”
Mythos debuted in April to widespread excitement as a model able to uncover code vulnerabilities that neither seasoned developers nor other AI systems could detect.
Also: US workers are the world’s biggest AI skeptics – and it’s not just about job loss
Positioned as the core of a Manhattan Project-style collaborative initiative known as Project Glasswing, Mythos was deemed far too risky to be widely accessible. Because of this, the model has only been offered to Glasswing partners, including Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks.
Until recently, Mythos was treated as a preview offering. Now, Anthropic is introducing Claude Mythos 5, which will be open to everyone who had access to Mythos Preview. The company stated, “We intend to gradually broaden availability through a more structured trusted-access program.”
No major details were shared about how Mythos Preview and Mythos 5 differ, aside from the latter appearing to be the official post-beta release of Mythos.
Also: Apple’s new Siri AI comes with hidden costs that power users should know of
Anthropic also launched Fable 5. The company explained that this technology runs on the same core model as Mythos, but with added safety layers. Those layers prevent the model from responding in certain high-risk domains of cybersecurity and biology.
Yes, the mention of “biology” is concerning. Have they actually encountered biological weapon-related prompts or outputs in their logs? I’ve reached out to Anthropic. I seriously doubt they’ll respond.
Notably, if a Fable 5 query drifts into one of those sensitive areas, the system automatically switches to Opus 4.8, which carries its own set of limitations. Starting with Opus 4.7, Anthropic has blocked “Activities that are almost always used maliciously and have little to no legitimate defensive application, such as mass data exfiltration or ransomware code development.”
Professionals who hold a security clearance from Anthropic can use Opus 4.7 and 4.8 to carry out restricted security tasks as part of their work. Disclosure: I’m an authorized participant in Anthropic’s Cyber Verification Program, so I have access to these capabilities for my cyberwarfare, cyberdefense, and counterterrorism work. It remains unclear whether those of us certified through the Cyber Verification Program will be able to run restricted queries on Fable 5.
Also: I used ChatGPT to build a free PDF editor because I didn’t trust it to change my files – it’s glorious
Anthropic appears confident that Fable 5 can’t be exploited for harmful purposes. They stated, “Early data indicates at least 95% of Fable sessions operate entirely on Fable’s own responses, without any fallback. We rigorously red-teamed our classifiers to verify their resilience against jailbreak attempts. Internally, we conducted an external bug bounty that yielded no universal jailbreaks after more than 1,000 hours of testing. We then collaborated with outside red-teaming organizations, which also failed to discover universal jailbreaks.”
From the announcement, there isn’t a lot of detail to pass along about Fable 5. However, Anthropic did share some early customer feedback about the release.
An unnamed representative from vibe-coding platform maker Base44 said, “Fable is far more capable and excels at building complete apps in a single attempt, and its tool calling is outstanding.”
Also: AI can identify intimate partner violence years before people disclose it, but is that safe?
Genspark offers an all-in-one AI workspace. A representative from the company said, “Fable ranked #1 in our evaluations, outperforming every model we tested. It was notably stronger on the most challenging tasks in the set — UI design and game development.”
A representative from e-commerce marketplace Rakuten said, “At the highest effort setting, Fable reviews and validates its own output. For us, that’s what enables highly autonomous operations — the additional reasoning more than justifies itself.”
On the topic of cost, pricing for Fable 5 and the new Mythos 5 is set at $10 per million input tokens and $50 per million output tokens. That’s approximately twice the cost of Claude Opus 4.8.
Anthropic has an unusual rollout strategy for Fable 5, based on their anticipation of strong demand. Here’s what the company outlined:
• From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no additional charge.
• On June 23, Fable 5 will be removed from those plans. Using it beyond that date will require usage credits.
• After that, we plan to bring Fable 5 back as a standard feature of subscription plans. We aim to do this as soon as possible.
Anthropic hasn’t offered any in-depth explanation for its naming choices for Mythos and Fable. According to Britannica, a mythos is a “complex body of sacred narratives, traditional stories, or belief systems that explain the origins of the world and cultural values.” In contrast, a fable is a “short, fictional story designed to teach a specific moral lesson.” Interpret that however you like.
I’m certain we’ll learn more about both of these releases. If my Max plan supports Fable 5, I’ll put it through its paces with some coding challenges once it’s available. So stay tuned.
Even at double the price of Opus, would you use Claude Fable 5 for coding work if it gave you Mythos-level capability with built-in safety fallbacks? Share your thoughts in the comments below.
You can keep up with my daily project updates on social media. Be sure to subscribe to my weekly newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.
