Anthropic releases new models, cost structures and safeguards
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Lindsey Bailey/Axios
Anthropic on Tuesday released upgraded versions of its most powerful AI models, cut costs for some users, adjusted safety interventions and rolled out new privacy safeguards for business users.
Why it matters: This could be the new playbook for a release: it's not enough to have a more intelligent model, you also need new tools and updates that address customer concerns.
Driving the news: Claude Fable 5.1 is now generally available, while the more restricted Claude Mythos 5.1 remains limited to vetted partners working in areas like cybersecurity and life sciences.
- Anthropic says Fable 5.1 improves on its June release across coding and multi-step knowledge work.
- Its base price remains $10 per million input tokens and $50 per million output tokens.
- Anthropic is also cutting the cost associated with Claude referencing or pulling on information it's already processed by 75%, which should make long-running AI agents cheaper to operate.
Zoom in: Anthropic is also introducing new enterprise safeguards that it says can provide protections equivalent to zero-data-retention policies while still allowing the company to police malicious use.
Zoom out: Anthropic said on X that it "improved our safeguards" which it defines as users being less likely to trigger safeguards that route them to more restricted responses.
- Medical or biology questions will have 85% fewer interventions, while some users could see roughly 60% fewer cybersecurity-related interventions per session.
- That addresses a major criticism developers have had regarding Anthropic. Routing company founder Joseph Perla tells Axios that his users were looking to be less reliant on Claude in part due to these interventions.
- Anthropic told Axios in June that the company was initially being more conservative on safeguards and they were working to improve on that over time.
Flashback: Fable 5 and Mythos 5 debuted in June as Anthropic's highest-capability models, with Mythos access deliberately restricted because of its advanced cyber and biological capabilities.
Between the lines: With one update, Anthropic unveiled new models, new safeguards, cybersecurity and intervention changes, cost cuts and added new enterprise IP protections.
- Competing AI lab OpenAI was rumored to release its latest model, Astra, earlier in the summer.
- The company paused some work on models following the HuggingFace incident.
What they're saying: "Releasing Fable 5.1 before Astra is quite a flex," Gavin Baker, managing partner and CIO at Atreides Management, said on X.
- That's perhaps even more true as it's widely expected that Anthropic could file IPO paperwork this month.
The bottom line: Anthropic has a lot of new things it's doing to retain customers.
