Anthropic has launched Claude Fable 5, its inaugural Mythos-class model available to the public.

      TL;DR: Anthropic has introduced Claude Fable 5, its first publicly accessible Mythos-class model, featuring safeguards to prevent high-risk queries and reverting to Opus 4.8 for safe responses. Released on Tuesday, Fable 5 is based on the same architecture as Anthropic's restricted Mythos system, marking the first time Mythos-class intelligence is available to the public. It is offered to enterprise clients and paid users, with new protections in place that inhibit responses related to cybersecurity, biology, chemistry, and what Anthropic refers to as “distillation” scenarios, defaulting to Claude Opus 4.8 instead. Initial data suggests these fallback situations occur in less than 5% of interactions, indicating that most users will engage with the complete Fable 5 model for the majority of their inquiries.

      Dianne Penn, Anthropic's head of product management for research, explained to CNBC that the focus is on a "race to the top," aiming to deliver the technology effectively while ensuring adequate safety measures to provide substantial benefits over potential harms. Penn described the model as a “significant jump” in capability from Opus 4.8, necessitating the extra safety protocols. According to Anthropic’s blog, Fable 5 outperformed Opus 4.8 by over 10% in certain benchmarks, although independent assessments have not yet been released.

      Additionally, Anthropic launched Claude Mythos 5, an updated version of the restricted Mythos model which does not include the cybersecurity and biology safeguards. Mythos 5 is still accessible only to about 50 vetted partners involved in Project Glasswing, Anthropic's cybersecurity initiative that has identified over 10,000 high or critical-severity vulnerability candidates in major software projects since April. Both models share the same foundational architecture, with Fable 5 incorporating the safety limitations deemed necessary for broader use.

      Pricing reflects the enhanced capabilities, with Fable 5 priced at $10 per million input tokens and $50 per million output tokens, double that of Opus 4.8, which was launched at $5 and $25 respectively. Until June 22, Fable 5 is available at no additional cost for Pro, Max, Team, and seat-based Enterprise subscribers, after which access will require usage credits.

      Penn noted that early users have reported improved cost efficiency per task, claiming that Fable 5’s advanced intelligence results in fewer iterations needed to complete tasks. “You just get a higher ROI by having more intelligent models,” she stated to CNBC. These efficiency claims are based on early feedback from users and haven't been independently verified.

      The launch is strategically timed for significant commercial impact. Last week, Anthropic confidentially submitted its IPO prospectus to the Securities and Exchange Commission, paving the way for a potential listing this fall. The company reported in May that its revenue run rate surpassed $47 billion, up from around $10 billion in annual revenue in 2025, and it recently secured a $65 billion Series H round at a valuation of $965 billion.

      This valuation has placed Anthropic ahead of OpenAI for the first time. OpenAI recently finished its funding round at $852 billion and also filed confidential IPO paperwork on Monday. SpaceX, which took over Elon Musk's AI startup xAI earlier this year, is expected to start trading on Friday at an estimated valuation nearing $1.8 trillion.

      This convergence means that three of the most valuable private tech companies will be entering the public markets within months of one another, and the fall 2026 IPO period could potentially yield more than $200 billion in new public market value from these listings alone. Claude Fable 5 serves as a showcase of Anthropic's ability to provide cutting-edge intelligence to customers at scale without the catastrophic misuse concerns that led to the restriction of Mythos. The company has implemented a 30-day data retention policy for all Fable 5 and Mythos 5 traffic for safety monitoring, emphasizing that this data will not be used for training.

      Whether a model that limits fewer than 5% of sessions can sufficiently mitigate the types of attacks that Mythos Preview was designed to detect remains an unanswered question by Anthropic. For investors assessing Anthropic’s IPO prospects, the Fable 5 launch represents a concrete response to one of the company’s key strategic challenges: how to monetize Mythos-class features without deploying the cybersecurity toolkit that grants the model both value and risk. The current solution appears to be a model that is significantly more intelligent everywhere else, at double the cost, and the upcoming quarters will reveal if this trade-off can generate sufficient revenue to validate a near-trillion-dollar public valuation.