Skip to content
News

Anthropic Claude Fable 5.1 Cuts Agentic AI Costs by 45%

Anthropic Claude Fable 5.1 Cuts Agentic AI Costs by 45%
Anthropic launches Claude Fable 5.1 and Mythos 5.1, with up to 45% lower costs for agentic tasks, refined safeguards, and improved data retention.

Anthropic has launched two new AI models, Claude Fable 5.1 and Mythos 5.1, positioning them as a response to customer concerns over pricing, data retention, and restrictive safeguards. The company says Fable 5.1 delivers stronger performance than its predecessor while costing around 25 percent less in typical use, and up to 45 percent less for complex agentic tasks.

The reduced pricing is attributed to lower charges on cached data that has already been processed and stored. This makes the model more efficient for developers and enterprises running demanding, multi-step workflows where costs can accumulate quickly.

Performance and early impressions

Early users have shared their experiences with the new model. Every chief executive Dan Shipper described it as the strongest coding model his team has used, adding that it is now fast, token-efficient, and communicates in a more natural manner. Box chief executive Aaron Levie reported that his company’s agent, running on Fable 5.1, picked up on subtleties and ambiguities in data that Fable 5 had missed in the same test.

On benchmarks, observers noted that Mythos 5.1 with low reasoning matched the score of the previous model set to maximum reasoning, suggesting notable efficiency gains between generations.

Safeguards, security, and data retention

Fable 5.1 introduces more precise safeguards that Anthropic says are less likely to block basic biology questions than Fable 5. Mythos 5.1, however, retains the same biology restrictions as the earlier model. The company also outlined progress on data retention, stating that its Enterprise Frontier Safeguards provide complete privacy by storing data on the customer’s own cloud servers rather than Anthropic’s, with a rollout beginning later in 2026.

On security, Anthropic said it is now permitting Fable 5.1 to be used for identifying software vulnerabilities. Certain cybersecurity tasks will still be redirected to Opus models, including penetration testing, exploit generation, and binary-based vulnerability scanning.

Claude Fable 5.1 is now available across all platforms. Mythos 5.1 is available only to participants in Project Glasswing.

Source
Image: theverge.com

The UK tech briefing

Smartphones, AI, computing and deals — the essential stories without the noise.

Mailing provider can be connected when your UK list is ready.

Shop on Amazon UK — Discover deals Shop on Amazon UK — Discover deals