Anthropic Debuts Claude Fable 5 and Mythos 5 With Massive Pe

Anthropic has launched Claude Fable 5, a high-performance AI model designed for general use that surpasses the capabilities of all its previous public releases. Announced on June 9, 2026, the new model arrives alongside Claude Mythos 5, a specialized version reserved for cybersecurity experts and infrastructure providers. Both models represent a significant leap in software engineering, vision, and complex reasoning tasks.

The Claude Fable 5 model is engineered to handle long, complex tasks with higher autonomy than its predecessors. In early testing, the model demonstrated its efficiency by helping Stripe migrate a 50-million-line Ruby codebase in a single day—a task that typically requires a full engineering team several months to complete. This level of performance is supported by a pricing structure of $10 per million input tokens and $50 per million output tokens, which is less than half the cost of the previous Claude Mythos Preview.

Safety remains a central focus for this release. Because Claude Fable 5 possesses advanced capabilities that could be misused in areas like cybersecurity, Anthropic has implemented conservative safeguards. These protections may occasionally affect harmless queries, redirecting them to the Claude Opus 4.8 model in roughly 5% of sessions. Meanwhile, Claude Mythos 5 offers the same underlying power but with lifted safeguards for trusted partners, currently deploying through Project Glasswing in coordination with the US Government to secure critical software infrastructure.

Beyond coding, the new models show state-of-the-art results in vision and memory. Claude Fable 5 can extract data from detailed scientific charts and even reconstruct web application source code from simple screenshots. In a demonstration of its raw processing power, the model successfully completed Pokémon FireRed using only visual input, whereas earlier versions required specialized helper tools to move through the game. The model also features persistent file-based memory, allowing it to maintain focus across millions of tokens during extended workflows.

Analytical performance has also seen a boost. On the Hebbia Finance Benchmark, Claude Fable 5 achieved the highest score of any tested model, showing particular strength in document-based reasoning and table interpretation. IMC reported that the model excelled in trading-analysis evaluations, covering everything from factual lookups to expected-value analysis. These improvements suggest that the model is better equipped for senior-level reasoning tasks that involve large amounts of data and multi-step problem solving.

The release of Claude Fable 5 is part of a broader effort to make advanced AI more accessible. Anthropic plans to expand access to the Claude Mythos 5 variant through a trusted access program in the near future. For now, general users can access the core capabilities through the Fable 5 model, which is available starting today. The company is also working to refine its safety filters to reduce false positives as more capable models are prepared for launch in the coming months.

While we strive for accuracy, bytevyte can make mistakes. Users are advised to verify all information independently. We accept no liability for errors or omissions.

Sources

Claude Fable 5 and Claude Mythos 5

AI-generated image.

✔Human Verified

Sources

Related Articles