Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on Monday, delivering a major upgrade to its top-tier AI model class just three months after launching Fable 5 with significant improvements in coding, scientific research, and long-running problem-solving tasks.
The two models share the same underlying architecture but differ in their safety guardrails. Fable 5.1 is generally available to all customers, while Mythos 5.1 is restricted to vetted participants in Anthropic trusted access programs for cybersecurity and life sciences work. The split represents a new approach to managing the risks of increasingly capable AI systems by offering different protection levels for different use cases.
Coding and Research Breakthroughs
Fable 5.1 sets new benchmarks across several key categories. On Terminal-Bench 4.0 for agentic coding, the model scored 55.8 percent compared to 42.0 percent for Fable 5 and 52.3 percent for Opus 5. Mythos 5.1 scored even higher at 60.9 percent on the same benchmark, reflecting the gap between the models on tasks where earlier cyber safeguards had intervened. On Terminal-Bench-Science 0.1 for agentic scientific research, Fable 5.1 reached 52.6 percent, more than doubling Fable 5 score of 24.7 percent and significantly surpassing Opus 5 at 29.0 percent.
The model also achieved strong results on Humanity’s Last Exam, reaching 60.9 percent without tools and 65.0 percent with tools, surpassing both Fable 5 and Opus 5. On CursorBench 3.2.0 for agentic coding, Fable 5.1 scored 73.4 percent, setting a new high-water mark for the category. On GDPval-AA v2 for knowledge work, the model scored 1853, ahead of Opus 5 at 1824 and Fable 5 at 1723.
One of the most notable real-world demonstrations came from investment firm Millennium, where Fable 5.1 identified the root cause of a rare crash in internal systems that had baffled human engineers and competing AI models for several years. The model disassembled an external vendor library, matched it against a core dump, and traced the crash to a bug in that library, according to Damien, a senior portfolio manager at the firm.
MongoDB staff engineer Ron Sanzone reported that Fable 5.1 built a complex prototype in about three days, doing initial research across all of the company services code and documentation to produce a novel and extensible design. The model then ran for hours unattended, with strong verification loops, to implement the entire prototype.
Major Cost Reductions for Developers
While baseline token pricing remains identical to Fable 5 at $10 per million input tokens and $50 per million output tokens, Anthropic significantly reduced the cost of cache reads, where the model reuses context that has already been processed and stored in memory. For standard API workloads, overall billing drops by roughly 25 percent. For highly agentic workflows where the model constantly re-reads long context windows during multi-step tasks, savings reach approximately 45 percent.
The cost reduction is particularly relevant for enterprise customers running complex, multi-step AI workflows. Cognition, the company behind the Devin AI coding agent, said it was moving its Opus 5 traffic to Fable 5.1 on launch day. It matched or edged out Fable 5 in our testing at a lower cost per task, and with the new cache read pricing a Fable-class model is finally economical for the workloads we had kept on Opus, said Walden Yan, Cognition co-founder and CPO.
Every CEO Dan Shipper described it as friendly Fable, noting that it was about twice as fast as Opus 5 and used half as many tokens, making it an obvious upgrade for anyone who used Opus as their daily driver.
Safety Improvements and Enterprise Features
Anthropic also introduced Enterprise Frontier Safeguards, a system that gives customers complete privacy equivalent to a zero data retention policy while maintaining strong protection against adversarial use. The system stores data in cloud infrastructure controlled entirely by the customer, not Anthropic, and will be made available in phases starting later this fall. Until EFS is available, eligible customers can use Fable 5.1 with zero data retention.
In cybersecurity, the updated safeguards block 60 percent fewer false positives than the previous version. This improvement is partly because Fable 5.1 can now be used to discover software vulnerabilities, though not to develop exploits for them. In biology, Anthropic established an access program in partnership with the US government to enable access to Mythos 5.1 advanced biology capabilities, with enrollment expected to open soon for scientists.
Jane Street Capital reported that in internal benchmarks, Claude Fable 5.1 solved more of their coding problems than Fable 5 or Opus 5, and achieved state of the art on trading intuition. Red Hat distinguished engineer Josh Boyer said that Fable 5.1 correctly identified the root cause of every broken build tested across all effort levels, communicating more effectively than earlier Anthropic models with updates that were more concise and easier to follow.
Availability and Next Steps
Fable 5.1 is available through Anthropic API as claude-fable-5-1, as well as through AWS, Google Cloud, and Microsoft Azure. Mythos 5.1 is available to vetted cyberdefenders through the Cybersecurity Verification Program and to life scientists through a government-coordinated access program. Anthropic said it is coordinating with the US government to expand access to Mythos 5.1 to a broader set of domestic and international partners as quickly as possible.
The release comes amid a rapid pace of model updates across the AI industry, with Anthropic pushing out Fable 5.1 just 90 days after Fable 5 launched in June. The company also noted it is continuing work to bring many of the improvements from Fable 5.1 to the rest of its model family, including smaller and more affordable tiers like Sonnet and Haiku.
discussion