Anthropic Unveils Claude Fable 5: Powerful but Pricey and Restricted
Anthropic has just launched Claude Fable 5, the first model in its new Mythos series, setting a new standard in AI performance while introducing stricter controls and higher costs. The model tops nearly every benchmark, including SWE-bench Verified at 95 percent, signaling a significant leap in coding and problem-solving capabilities. However, this power comes at a premium—subscription fees are double those of its predecessor, Opus 4.8, priced at $10 or $50 per million tokens depending on usage tiers.
A New Tier with Heavy Guardrails
Claude Fable 5 introduces a 30-day data retention policy, a shift from Anthropic’s previous zero-retention stance, even for users under zero-data contracts. This move aligns with growing regulatory and ethical demands but may raise concerns among privacy-focused developers. Additionally, the model enforces strict safety filters, blocking around 9 percent of requests—a figure notably higher than typical industry averages. While this reduces risks of misuse, it could limit creative or unconventional applications that fall outside predefined boundaries.
Performance vs. Practicality
Despite its cutting-edge performance, the model’s high cost and restrictive policies may deter some users. Enterprises and researchers will need to weigh the benefits of top-tier benchmark results against budget constraints and data governance requirements. Whether the Mythos series becomes a game-changer or a niche tool remains to be seen, but its launch underscores Anthropic’s commitment to balancing innovation with safety—even at a premium.
Source: The Decoder. AI-assisted editorial synthesis — TechnoExpress.