Anthropic Slashes Prices as Agentic AI Race Heats Up
AI-generated image · US National Wire
The launch of Fable 5.1 brings significant cost reductions for complex tasks and new privacy controls, but skepticism remains over the sustainability of the AI pricing war.
Anthropic has released Fable 5.1 and Mythos 5.1, updated versions of its advanced AI models designed to address customer complaints regarding pricing, data retention, and overly restrictive safeguards.
According to reporting from The Verge, Anthropic claims Fable 5.1 is typically 25 percent cheaper than its predecessor, Fable 5. For complex agentic tasks, the company says the model is up to 45 percent less expensive, a reduction driven by lower pricing for cached data that has already been processed and stored.
Early adopters have praised the performance of the new release. Every CEO's Dan Shipper described Fable 5.1 as the strongest coding model they have used, noting it is fast and token-efficient. Box CEO Aaron Levie stated that his company's agent using Fable 5.1 successfully identified ambiguities and subtleties in data that Fable 5 had previously missed.
Beyond pricing, Anthropic is shifting its approach to data privacy. TechCrunch reports that the company is embracing zero data retention, which allows clients to operate models on their own infrastructure. A high-privacy service titled Enterprise Frontier Safeguards will begin rolling out this fall, storing data on customer cloud servers rather than Anthropic's own. While Anthropic will still monitor for human or agent misuse, TechCrunch notes that clients will maintain control over how that monitoring is conducted.
On the technical side, TechCrunch reports that the new models set records on benchmarks such as Humanity’s Last Exam for general reasoning and Terminal-Bench 4.0 for coding via CLI. Anthropic also highlighted scientific outputs from the models, including a high-resolution map of Venus and a custom GPU optimization.
However, the updates come with caveats. The Verge reports that Lisan al Gaib noted Mythos 5.1 with low reasoning scores the same as the previous model set to Max reasoning. Additionally, while Fable 5.1 features more precise safeguards to avoid blocking basic biology questions, The Verge notes that Mythos 5.1 retains the same biology restrictions as the prior version.
Anthropic is also expanding the utility of Fable 5.1 to include the identification of software vulnerabilities. However, the company will continue to route specific cybersecurity tasks—such as binary-based vulnerability scanning, exploit generation, and penetration testing—to Opus models.
Availability varies by model: Fable 5.1 is available on cloud platforms and via the Anthropic API, while Mythos 5.1 is restricted to Project Glasswing participants and registered partners in life sciences or cybersecurity research.