Anthropic launches Opus 5
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Anthropic's newly launched Opus 5 outperforms the larger Fable 5 on multiple benchmarks at a lower cost, triggering safety classifiers 85% less often and exempting users from the restrictive 30-day data retention policy. This allows production-grade coding and reasoning agents to handle sensitive workloads with fewer compliance-related blockers, while a new Automatic Fallbacks feature eliminates API downtime by silently routing safety-blocked prompts to smaller models instead of throwing errors.
Opus 5 outperforms the larger Fable 5 on several benchmarks while costing less and triggering safety classifiers 85% less often, and it's exempt from the 30-day data retention policy—making it the default choice over Fable for most production workloads. Its standout gain is self-verification and iterative retry (it built a CV pipeline from an incomplete prompt), and the new opt-in Automatic Fallbacks routes blocked requests to a weaker model instead of erroring, so agent pipelines won't hard-fail on classifier hits. Note the asymmetric cyber safeguards: source-code vuln search is allowed but binary scanning is blocked, which will break some security tooling.