
What happened
Anthropic released Opus 5 on Friday, a new model that is smaller than Fable 5 but outperforms it on several benchmarks included in the announcement. The model comes only two months after Opus 4.8 launched on May 28.
Why it matters
Opus 5 is both cheaper and subject to fewer restrictions than Fable 5, making it likely preferable for most use cases. Unlike Fable and Mythos, Opus 5 is not subject to the 30-day data retention policy, addressing privacy concerns some users had raised. Anthropic emphasized in its announcement that Opus 5 is "much stronger at verifying its work and iterating carefully until it succeeds," citing examples like writing its own computer vision pipeline from an incomplete prompt.
What to watch
Anthropic is rolling out Automatic Fallbacks, a beta feature that routes requests to a less powerful model when a prompt triggers the safety classifier, giving users a functional response instead of an error. Anthropic expects these safeguards to engage 85% less often for Opus 5 than for Fable 5. Haiku remains the only model in the 5 series still waiting for an upgrade.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
Opus 5 represents Anthropic's continued cadence of releasing new model versions across its product line. The launch comes just two months after Opus 4.8 became available on May 28, and follows the June release of Mythos 5, Fable 5, and Sonnet 5. With Haiku still awaiting a 5-series update, Opus 5 fills a gap in Anthropic's lineup by offering better performance than Fable 5 at lower cost and with fewer restrictions, a positioning that Anthropic suggests will make it preferable in most use cases.
The removal of the 30-day data retention policy represents a meaningful shift in Anthropic's approach to safety and privacy. The retention requirement on Fable and Mythos had raised concerns among privacy-conscious users, and Opus 5's exemption from this policy addresses that friction point directly. At the same time, Anthropic has not abandoned safety entirely—safeguards remain around cybersecurity tasks, though the company expects them to engage 85% less often for Opus 5 than for Fable 5, reflecting a more permissive approach to a more capable model.
The introduction of Automatic Fallbacks signals an evolution in how Anthropic handles the tension between safety and usability. Rather than returning an error when safeguards trigger, the feature gracefully degrades to a less powerful model, ensuring users get a functional response. This approach aims to reduce friction for API users while maintaining safety boundaries.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Anthropic CEO Dario Amodei published an essay over the weekend proposing "pacing the frontier" and slowing imp…

Anthropic CEO Dario Amodei published a 3,800-word essay Saturday calling for slowing the pace of AI model capa…

Sam Altman called for a uniform, nationwide safety framework for advanced AI and said OpenAI now sets explicit…

At the Fortune Leaders Forum in Macau on Sept

NC State first-year Amanda Cullen says most of her classes ban AI unless cited, and questions whether outright…

A Stony Brook University-led study of 14,419 self-published genre-fiction books on Amazon found 2,880, about 2…
