Anthropic has paused access to Claude Fable 5 and Mythos 5 after rolling them out on June 12, 2026. Fable 5 is their top-tier general model, outpacing previous releases on benchmarks for coding, complex reasoning, vision tasks and scientific research. To prevent misuse—particularly in cybersecurity—queries on sensitive topics silently revert to Claude Opus 4.8. Those safeguards trigger in under 5 percent of sessions but will tighten further as Anthropic refines its filters. Separately, Mythos 5 ships without those lift­ed cybersecurity restrictions for vetted cyberdefenders via Project Glasswing, with broader trusted access coming later. Both charge $10 per million input tokens and $50 per million output tokens, less than half of the prior Mythos Preview price. In software engineering tests, Stripe used Fable 5 to migrate a 50-million-line Ruby codebase in a single day—a task that would have taken a team two months. On Cognition’s FrontierCode suite, Fable 5 tops all frontier models for generating production-quality code. In knowledge work, it leads Hebbia’s Finance Benchmark for senior-level reasoning and aced IMC’s trading-analysis challenges. Vision tasks also shine: Fable 5 can pull exact figures from scientific charts and reconstruct web-app source code from screenshots with minimal guidance. Its long-context memory spans millions of tokens, boosting performance in tasks like playing Slay the Spire threefold compared to Opus 4.8. Mythos 5 powers internal protein-design experiments that run ten times faster than before. In one blind comparison, it matched or beat expert scientists across all protein-engineering steps and produced nine strong drug candidates now under investigation. For genomics, Mythos 5 spent a week autonomously building a dataset of single-cell profiles from 138 species, then trained a compact model that outperformed a recent Science publication. Early alignment tests show Mythos 5’s misaligned behavior stays low and on par with Opus 4.8. Users report it handles long-horizon coding problems and agentic tasks more reliably than any previous Claude release.