Anthropic today rolled out Claude Fable 5, the newest Mythos-class model, with stronger capabilities and new safety controls — and it’s immediately available to users via Claude products and the Claude API (API id: claude-fable-5). Anthropic says Fable 5 outperforms prior Claude releases on software engineering, knowledge work, vision, and scientific research — with its biggest advantage coming on longer, more complex tasks. The company claims the model can operate autonomously for extended periods and maintain focus across millions of tokens, making it well suited to long-running workflows.
Real-world results cited by Anthropic include a notable software-testing win: Stripe reported that Fable 5 completed a large Ruby migration in one day, a job Stripe estimates would have taken a team more than two months by hand. Anthropic also highlights strong performance on finance, memory, and scientific tasks.
Because of the model’s power, Anthropic is deploying extra safeguards. Some requests to Fable 5 — notably certain cybersecurity, biology, chemistry, and “distillation”-related queries — will automatically fall back to Claude Opus 4.8. Anthropic says these fallbacks activate in under 5% of sessions on average; in other words, over 95% of Fable sessions run with no fallback. The company acknowledges the safeguards were tuned conservatively and may catch harmless requests, and it plans to reduce false positives after launch.
New protective systems include classifiers to detect potential misuse and jailbreak attempts; when triggered, these systems prevent Fable 5 from responding directly to flagged queries. The cybersecurity controls specifically target exploitation and other offensive cyber tasks, while biology and chemistry guardrails address dual-use risks that could enable harm.
Alongside Fable 5, Anthropic launched Claude Mythos 5 for a smaller, approved cohort of users. Mythos 5 shares the same underlying model as Fable 5 but is being distributed initially through Project Glasswing in cooperation with the U.S. government to cyberdefenders and providers of critical software infrastructure. For approved cybersecurity users, Mythos 5 will lift some safeguards. Anthropic also plans a trusted-access program for selected biology researchers and companies.
Business traffic on Mythos-class models will be retained for 30 days, Anthropic says, and that data will not be used to train future Claude models. Pricing is set at $10 per million input tokens and $50 per million output tokens. Fable 5 is included on Pro, Max, Team, and seat-based enterprise plans through June 22; from June 23 onward, access will require credits unless Anthropic extends capacity.
What this means for crypto and blockchain teams: models that can sustain context over “millions of tokens” and handle extended autonomous work could speed source-code migrations, large-scale audits, ledger analysis, and long-form research workflows. At the same time, Anthropic’s conservative safeguards and the separate Mythos track signal that capabilities useful for defensive security work are being guarded to reduce misuse risk.
Developers can access Fable 5 now via the claude-fable-5 API identifier or through the hosted Claude products.
Read more AI-generated news on: undefined/news
