Anthropic Unveils Claude Fable 5: A Safer AI Experience for All

TL;DR

• Anthropic has launched Claude Fable 5, the first publicly accessible version of its Mythos-class model, with availability through the Claude API and select subscription plans.
• The model includes hard safety guardrails that block or redirect risky requests in areas like cybersecurity and biology to Claude Opus 4.8.
• Anthropic is pricing Fable 5 at $10 per million input tokens and $50 per million output tokens, while rolling out broader access in stages.

Anthropic has opened a new chapter in its AI lineup with the launch of Claude Fable 5, the first publicly available model from its Mythos-class family. The release marks the company’s latest effort to balance frontier-level capability with stronger safety controls, especially in domains where powerful AI systems could be misused.

A public debut for Anthropic’s most advanced model tier

Fable 5 is being positioned as Anthropic’s most capable model for broad use, with reported strengths in software engineering, knowledge work, and vision tasks. According to the company, it is now available through the Claude API and consumption-based Enterprise plans, with temporary no-cost access also rolling out to Pro, Max, Team, and seat-based Enterprise users through June 22.

Anthropic says the model is the public-facing version of its Mythos technology, which had previously been restricted to private testing and controlled access. That framing reflects the company’s strategy of separating raw capability from the permissions granted to end users.

Safety guardrails at the center of the launch

The defining feature of Fable 5 is not just its performance, but its guardrails. In high-risk areas such as cybersecurity, biology, chemistry, and distillation, the model is designed to block direct responses and instead fall back to Claude Opus 4.8. CNBC reported that Anthropic said these safeguards were essential to making a broader release possible.

BBC reported that Anthropic acknowledged the risks of releasing a model with this level of capability, and said the public version would ship with protective measures and user restrictions. In effect, Fable 5 is meant to deliver advanced reasoning and productivity benefits while reducing the chance of harmful assistance in dual-use domains.

Early testing suggests the protections are holding

Anthropic says it spent more than 1,000 hours of red-team testing trying to find ways around the model’s safety systems. The company said it did not find a universal jailbreak, and TechCrunch reported that at least 95% of Fable sessions are handled entirely by the model without needing to defer to Opus 4.8.

That matters because it suggests the restrictions are not dominating the product experience in ordinary use, even though they remain firm in sensitive contexts. For Anthropic, this is the key argument for the launch: frontier models can be made more available without abandoning strict controls.

Limited access, premium pricing

The rollout is being staged carefully. Anthropic said access will expand across paid subscription tiers through June 22, after which Fable 5 will move to usage-credit pricing on those plans until a standard subscription arrangement is restored. For enterprise customers, the model is also available through API and consumption-based plans.

Pricing is set at $10 per million input tokens and $50 per million output tokens, which Anthropic says is double the cost of Claude Opus 4.8. That makes Fable 5 one of the company’s most expensive offerings, reflecting both its advanced capability and the added operational burden of safety filtering.

Data retention becomes part of the tradeoff

Anthropic also said the launch of Fable 5 and Mythos 5 will require 30-day data retention on all traffic, even for enterprises that previously had zero-retention agreements. That policy shift underscores how tightly the company is controlling the deployment of its newest model tier.

For some customers, that may be a significant change in how they evaluate adoption, especially in regulated industries or sensitive research settings. The company appears to be betting that the combination of stronger performance and safety controls will outweigh those concerns for enough users.

What this means for the AI race

The launch puts Anthropic in a sharper position within the competitive frontier-model race. While rivals continue pushing for more capable systems, Anthropic is emphasizing a different differentiator: not just more power, but power delivered with deliberate limits.

That approach could become increasingly important as AI systems move into areas where misuse risks rise alongside utility. Fable 5 suggests Anthropic believes the next phase of AI adoption will depend not only on what models can do, but on what they are prevented from doing.

The bigger picture

The release of Claude Fable 5 also signals a broader shift in how advanced models may reach the public. Instead of waiting for a single “all-access” release, Anthropic is introducing layered availability, restricted capabilities, and fallback behavior as core product features.

For users, that may mean a more cautious AI assistant in high-risk scenarios, but one that is still highly capable for everyday coding, analysis, and productivity work. For the industry, it is another sign that safety engineering is becoming just as important as model scale in defining the next generation of AI products.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.