Anthropic rolled out Claude Fable 5 this week, positioning it as the most powerful AI model the firm has ever made broadly accessible. The company highlighted the model’s strong performance across software engineering, knowledge‑intensive work and visual tasks, noting that its advantage over rival systems widens as tasks become longer and more intricate.
Claude Fable 5 represents the first wide‑scale release from Anthropic’s Mythos family. Earlier, the company kept the class under wraps, citing concerns that its capabilities in cybersecurity could be misused if left unchecked. New safety mechanisms now permit a public offering. Those safeguards automatically block responses in specific high‑risk domains—chiefly cybersecurity and biology—areas widely regarded as sensitive for advanced AI. When a request hits a blocked topic, the system defers to Claude Opus 4.8, a model Anthropic praised for its “honesty” when it debuted last month.
In internal testing, Anthropic reported that 95 percent of Claude Fable 5 sessions completed without invoking the Opus fallback, suggesting the safeguards rarely interfere with ordinary use. The firm also introduced Claude Mythos 5 alongside the Fable release. While both share the same underlying architecture, Mythos 5 lifts certain safeguards, offering a less constrained experience for qualified users.
Access to the new models remains limited. Anthropic continues to grant entry through its private Project Glasswing initiative, a trusted‑access program that currently serves a growing roster of organizations. Those participants can upgrade to Mythos 5, and Anthropic says it plans to broaden the program over time, moving toward a more systematic rollout.
The pricing structure for Claude Fable 5 and Claude Mythos 5 marks a notable increase over the company’s previous flagship offering, Claude Opus 4.8. Users will pay $10 for every million input tokens and $50 for every million output tokens. Although these rates double the cost of Opus 4.8, they remain half of what Anthropic charges for the Mythos Preview tier.
Anthropic declined to comment on how the new models relate to the Claude Mythos Preview or why the “5” designation appears without any publicly released predecessors. The company’s blog post emphasized the technical merits of the release and outlined its plans to expand access, but offered few additional details about the models’ architecture or future roadmap.
Industry observers see the launch as a significant step for Anthropic, which has positioned itself as a safety‑first AI developer. By coupling a high‑performing model with robust safeguards, the firm aims to address growing concerns about AI misuse while still delivering powerful tools for enterprise and research customers.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.