previously, we reported that company a officially launched two cutting-edge large-scale models specifically designed for programming—claude fable 5 and claude mythos 5. these models not only deliver breakthrough capabilities in code generation, comprehension, and optimization but also feature a deeply integrated multi-layered security framework, aiming to mitigate cybersecurity threats arising from malicious exploitation at the source. concurrently, company a has overhauled its commercialization strategy: instead of including fable 5 and mythos 5 within its existing subscription-based service model, it has transitioned entirely to an api‑based pay-per‑call pricing structure, requiring developers to directly invoke the apis and pay for the actual computational resources consumed.
the current subscription‑based trial access will end on june 22, 2026:
at this stage, developers can still gain limited access to fable 5 and mythos 5 through subscription plans such as claude pro or max during the trial period, with calls prioritized against the allocated weekly quota of 5 hours. however, given these models’ extremely high computational demands, standard subscription quotas often deplete rapidly, leaving most users ultimately needing to switch to the api‑based pay‑per‑use model to ensure uninterrupted development.
effective june 22, 2026, these advanced models will be completely removed from all individual subscription plans. going forward, individual developers will only be able to access them via the api, billed according to actual token consumption; enterprise users may opt for the claude enterprise subscription plan, which employs a dual‑track pricing mechanism—“seat fees plus separate billing based on compute usage per seat”—with no free allowances and no bundled general‑purpose quotas.
this purely api‑based access policy is intended as a temporary measure. as fable 5 and mythos 5 are still in their initial release phase, company a has not yet opened subscription channels for these models. with the gradual launch of additional advanced models and improvements in infrastructure maturity, it is anticipated that these models will eventually be incorporated into the standard claude subscription lineup, allowing users to benefit from more cost‑effective plans and significantly reducing per‑call expenses.
all request data will be retained uniformly for 30 days for security auditing:
company a has simultaneously upgraded its enterprise‑level data governance strategy: regardless of whether requests originate from company a’s official api or are routed through third‑party platforms, all raw traffic data associated with interactions involving fable 5/mythos 5—including input prompts, output responses, and metadata—will be encrypted and stored for 30 calendar days, exclusively for security compliance reviews and tracing novel attack patterns. company a explicitly commits that such data will never be used in any model training processes. based on newly identified abuse patterns uncovered during audits, defensive rules will continue to be updated and integrated into the model’s security layer, dynamically enhancing its ability to detect and block high‑risk activities such as black‑and‑gray‑market toolchains, automated attack scripts, and jailbreak attempts, thereby effectively mitigating systemic industry risks stemming from technology misuse.
