Anthropic has introduced two new artificial intelligence models, Claude Fable 5.1 and Claude Mythos 5.1, around three months after releasing Fable 5. The company says the new models offer major improvements in coding, research and scientific work.
Fable 5.1 and Mythos 5.1 are built on the same core technology, but they use different safety systems. Fable 5.1 is available to all users, while Mythos 5.1 is limited to approved organisations working in cybersecurity and life sciences.
Anthropic places the new models above its existing Haiku, Sonnet and Opus series, which are designed for different levels of performance. According to the company, Fable 5.1 is its strongest model so far for coding, reasoning and knowledge tasks.
In the Terminal-Bench-Science 0.1 evaluation, Fable 5.1 scored 52.6%. This was higher than Fable 5 at 24.7%, Opus 5 at 29.0% and rival model GPT-5.6 at 22.4%.
The company also reported strong results in Terminal-Bench 4.0. Fable 5.1 achieved 55.8%, while Mythos 5.1 scored 60.9%. By comparison, Fable 5 recorded 42.0%, Opus 5 reached 52.3% and GPT-5.6 scored 37.3%.
Anthropic has also reduced operating costs. The price of cache reads, which allow models to reuse previously processed information, has been cut by 75% to $0.25 per million tokens. The company says this makes Fable 5.1 about 25% cheaper than Fable 5 in normal use and up to 45% less expensive for demanding tasks such as complex coding. Prices for standard input and output tokens remain unchanged.
The company announced a new privacy system called Enterprise Frontier Safeguards (EFS). This system will allow business customers to keep sensitive data on their own servers instead of Anthropic’s infrastructure. EFS is expected to roll out in phases from autumn 2026 across services including Claude Code, Claude Enterprise, Amazon Bedrock, Google’s Agent Platform and Microsoft Foundry.
Anthropic says it has also improved safety measures. The company reported a 60% reduction in false cybersecurity alerts. Fable 5.1 can identify software vulnerabilities, but Anthropic says it is not designed to exploit them.
The company highlighted several scientific applications of the new models. In drug discovery tests, Mythos 5.1 designed molecules for potential medicines and achieved a success rate of nearly 50% across 12 targets. Anthropic said this compares with an industry average of around 10% to 15%.
Fable 5.1 was also used to create a detailed map of part of Venus using older NASA radar data. Anthropic says the resulting map contained more detail than earlier versions.
In biology research, Mythos 5.1 helped some programmes run up to 2.5 times faster while reducing computing costs by 30% to 60%, according to the company.
Both models underwent safety testing related to chemical, biological and cybersecurity risks before launch. Anthropic said the tests did not find serious ways to bypass its safeguards.
Fable 5.1 is now available through Amazon Web Services, Google Cloud, Microsoft Azure and Anthropic’s API platform. Mythos 5.1 remains limited to approved organisations in the United States, although Anthropic says it is working with the US government to expand access in the future.
For the latest updates, follow News on our website
