Anthropic unveils Claude Fable 5.1 and Mythos 5.1 for coding and knowledge work

Anthropic, an artificial intelligence firm, has unveiled two new AI models, Claude Fable 5.1 and Claude Mythos 5.1, designed for coding and knowledge work. These models, according to Anthropic, are the world's most advanced despite being based on the same underlying architecture, differing in operational safeguards. Claude Fable 5.1 is available to all users, while Claude Mythos 5.1 is restricted to trusted access programs focusing on cybersecurity and life sciences research.

Anthropic highlighted that Claude Fable 5.1 outperforms its predecessor, Fable 5, in long-running problem-solving and technical tasks. It performs comparably or better than Fable 5 at reduced computational costs, especially when operated under Low or Medium effort parameters. This model will be 25% cheaper than Fable 5 for typical workloads billed by token due to reduced pricing on cache reads. For highly agentic tasks, savings could be up to 45%.

In a practical evaluation by investment firm Millennium, Claude Fable 5.1 was able to identify the source of an intermittent internal crash that had eluded human engineers and previous models for years.

Regarding data privacy, Anthropic introduced Enterprise Frontier Safeguards (EFS), which keeps customer data within customer-managed cloud infrastructure instead of Anthropic's servers. EFS for enterprise accounts will be rolled out in phases starting later this fall, with zero data retention as an interim provision for eligible users.

In terms of cybersecurity, the new safeguards block 60% fewer false positives compared to previous versions. This is partly because Claude Fable 5.1 can now help discover software vulnerabilities, though not develop exploits for them.

Anthropic also announced a reduction in the price of Fable 5.1's cache reads, which is where the model reuses previously processed context. This change has led to a 75% decrease in cache reads' cost, or $0.25 per million tokens, resulting in a 25% reduction in overall costs for typical workloads compared to Fable 5. The pricing remains consistent with Fable 5 at $10 per million input tokens and $50 per million output tokens.

Finally, Anthropic stated that they are continuing efforts to bring the improvements of Fable 5.1 to the rest of their model family. Currently, Claude Mythos 5.1 is available to selected US organizations, primarily cybersecurity experts and life scientists, with plans to expand access to a broader range of domestic and international partners.