Anthropic's 'Fable 5' AI Makes Gaming Dreams a Reality: Public Access Now Available!
Anthropic has launched Claude Fable 5, its powerful Mythos AI model, to the general public, featuring advanced capabilities in software engineering and knowledge work with strict safety guardrails. Rigorous testing by researchers like Ethan Mollick demonstrates its remarkable ability to generate complex applications and video games from single prompts. Despite its premium pricing, Fable 5 sets a new standard for accessible, high-performance AI.
Anthropic has officially launched Claude Fable 5, making its most powerful AI model, Mythos, available to the general public for the first time. This release marks a significant step, as Fable 5 is touted for its excellence in software engineering, knowledge work, and vision, though it comes equipped with stringent safety guardrails.
Initially, the Mythos model was launched as a preview in April, with access restricted to a select few partners due to paramount cybersecurity concerns. This access was later expanded to hundreds of organizations across 15 countries, primarily focusing on entities managing critical infrastructure. Now, a version of this advanced technology is accessible to anyone via Anthropic's Claude API and consumption-based Enterprise plans. Subscription access for Fable 5 will roll out in stages, included at no extra cost in Pro, Max, Team, and seat-based Enterprise plans until June 22. From June 23, it will require usage credits, with plans to restore it as a standard subscription feature as soon as feasible. Concurrently, Anthropic is deploying Mythos 5, an updated version, to already approved organizations.
A core aspect of Fable 5's release is its emphasis on safety. In high-risk domains such as cybersecurity, biology, chemistry, and distillation, the model is programmed to block responses and default to Claude Opus 4.8. This cautious approach is part of Anthropic's broader strategy, which includes a plea for major global AI labs to establish a coordinated 'brake pedal' on frontier AI development, fearing that systems could soon achieve recursive self-improvement (RSI) without human intervention. To mitigate risks, Anthropic rigorously stress-tested Fable 5's classifiers with numerous 'jailbreak' attempts. Internal and external bug bounties, involving over 1,000 hours of testing and external red-teaming organizations, reportedly found no universal jailbreaks, though the company acknowledges the potential for novel attacks.
In light of these safety considerations, Anthropic has implemented a mandatory 30-day data retention policy for all traffic with Fable 5 and Mythos 5, even for enterprises that previously had zero-retention agreements. The company explicitly states this data will not be used for training but solely to "defend against complex and novel attacks, including new jailbreaks," and "identify and reduce false positives." This policy could establish an industry precedent where access to increasingly powerful AI models is linked to mandatory data-retention for safety measures.
Despite these guardrails, Fable 5 demonstrates impressive capabilities. Anthropic reports that instances where Fable defers to Opus 4.8 are rare, with early data indicating that at least 95% of Fable sessions run entirely on its own responses. Third-party testing further validates its prowess: analytics company Hex noted that Fable was the first to achieve a 90% score on its core analytics benchmark for complex, long-running analytical tasks, showcasing "strong judgement and attention to nuance." Vibe-coding platform Base44 highlighted Fable's ability to excel at "one-shotting full apps" and its excellent tool-calling. AI-powered workspace Genspark also reported that Fable surpassed all other models in its evaluations, performing significantly better on tasks like UI design and game coding.
Renowned AI researcher Ethan Mollick from the University of Pennsylvania has extensively tested Fable 5, reporting that it consistently "outperformed basically every other public model I have used by a considerable margin." Mollick found it capable across many problems, producing "startling results" and working up to a dozen hours on multi-page specifications. Strikingly, he used Fable to create various video games—including a Pac-Man-like 'Snake', a subterranean exploration game 'Strata' with Myst-like graphics, and 'Duino', based on Rainer Maria Rilke's 'Duino Elegies'—all from a single initial prompt in Claude Code. Furthermore, Mollick successfully generated a highly accurate and detailed isochronic map. These demonstrations vividly illustrate Fable's potential to spin up complex software projects, like games and mapping tools, that previously required entire teams, from a single prompt.
However, accessing these advanced capabilities comes at a premium. Pricing for both Fable 5 and Mythos 5 is set at $10 per million input tokens and $50 per million output tokens, effectively doubling the cost of Opus 4.8. This pricing strategy may act as a deterrent for widespread adoption, as many enterprises are becoming critical of escalating AI costs. Despite this, some, like shopping rewards platform Rakuten, believe the benefits outweigh the expense, stating, "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." Anthropic anticipates high and unpredictable demand for Fable 5, acknowledging its launch alongside the company's preparations to enter public markets.