Claude Fable 5 brings Mythos power with safety clamps

Anthropic is rolling out Claude Fable 5 for general users as a “Mythos-class model made safe for general use,” pairing Mythos-level capability with safeguards that block high-risk cybersecurity and biology requests. Pricing sits at $10 per million input tokens

By the time Mythos showed up, it already carried a warning label: this was an unusually powerful model, and it was restricted for a reason. Now Anthropic is trying to give that same kind of capability to regular users—without handing them the dangerous edges.

Claude Fable 5 is the new release. Anthropic describes it as “a Mythos-class model made safe for general use.” It uses what the company says is the same underlying model as Mythos. but with safeguards attached. Those safeguards block responses in specific high-risk areas of cybersecurity and biology.

If a Fable 5 prompt veers into one of those high-risk areas, the model doesn’t keep going. Instead, it “drops down” to Opus 4.8, which has its own restrictions. Ever since version Opus 4.7. Anthropic blocks “Activities that are almost always used maliciously and have little to no legitimate defensive application. such as mass data exfiltration or ransomware code development.”.

There’s a real-world reason this matters: Mythos itself was launched in April to major fanfare as a model capable of finding vulnerabilities in code that neither experienced developers nor other AIs could find. It was also framed from the start as too dangerous to land in the wrong hands. That’s why Anthropic built Mythos into a Manhattan Project-like effort called Project Glasswing and limited it to Glasswing partners.

Those partners include Amazon Web Services. Anthropic. Apple. Broadcom. Cisco. CrowdStrike. Google. JPMorganChase. the Linux Foundation. Microsoft. Nvidia. and Palo Alto Networks. Up until now, Mythos has been treated as a preview product. Anthropic says it is now launching Claude Mythos 5. which will be available to all users who have access to Mythos Preview. Anthropic also said. “We plan to expand access over time through a more systematic trusted-access program. ” and it did not provide significant details about differences between Mythos Preview and Mythos 5 beyond the suggestion that Mythos 5 appears to be the post-beta version.

Fable 5 and Mythos 5 are being priced in a way that signals Anthropic isn’t trying to make this cheap. Pricing for Fable 5 and the new Mythos 5 is $10 per million input tokens and $50 per million output tokens—about twice the price of Claude Opus 4.8.

That cost becomes part of the rollout story. From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost. On June 23, Anthropic will remove Fable 5 from those plans. After that point. using Fable 5 requires usage credits. with the company aiming to restore Fable 5 as a standard part of subscription plans “as quickly as we can.”.

Anthropic’s safety claims come with a very specific kind of confidence. The company said. “Early data shows at least 95% of Fable sessions run entirely on Fable’s own responses. with no fallback.” It says classifiers were extensively red-teamed to test robustness against jailbreaks. and that an internal external bug bounty produced no universal jailbreaks in over 1. 000 hours of testing. It also says it worked with external red-teaming organizations, which also failed to find universal jailbreaks.

The tension in that promise isn’t just whether safeguards work—it’s what happens when they catch the wrong edge. One of the areas Anthropic flagged for blocking is “biology. ” and that detail was enough to raise red flags in the reporting around the launch. The central question is simple: have they seen biological weapon prompts or responses in their logs?.

There’s also a different layer for professionals with specialized access. Professionals with a security clearance from Anthropic can use Opus 4.7 and 4.8 to perform blocked security activities as part of their jobs. It’s not yet clear whether people certified through Anthropic’s Cyber Verification Program can perform blocked queries with Fable 5.

What do people actually say the model can do? The launch comments lean hard on usefulness for coding and creation.

A representative of Base44. an “vibe-coding” platform maker. said. “Fable is much deeper and better at one-shotting full apps. and its tool calling is excellent.” A representative from Genspark. an all-in-one AI workspace. said. “Fable came out #1 on our evals. winning head-to-head against every model we tested. It was significantly stronger on the hardest tasks in the set — UI design and game coding.” An e-commerce marketplace Rakuten representative said. “At the highest effort. Fable reflects on and validates its own work. For us, that’s what makes highly autonomous operations possible — the extra thinking pays for itself.”.

Taken together, the messaging paints a picture Anthropic wants to manage carefully: Mythos-level capability for everyday users, but with safety fallbacks and staged pricing that let the company control demand while it expands access.

If your subscription can host Fable 5—at least during the window where it’s included through June 22—the immediate choice may come down to the same question everyone seems to be circling: would you pay twice the Claude Opus 4.8 rate for coding work if the power came with built-in guardrails that can pull the model back to Opus 4.8 when it detects high-risk directions?.

The model is new, the rollout is temporary, and the whole point is that Mythos was never meant to be broadly available in the first place. Now Anthropic is testing what happens when that capability is loosened—just enough, and not all the way.

Anthropic Claude Fable 5 Claude Mythos 5 Mythos Opus 4.8 AI safety cybersecurity safeguards biology safeguards Project Glasswing pricing tokens Pro Max Team Enterprise usage credits