Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1, describing the pair as its most capable models yet for coding and knowledge work. Fable 5.1 is available to the public now, while Mythos 5.1 remains limited to organisations enrolled in Anthropic’s trusted access programme.
The two models share the same underlying system but differ in the safeguards applied to each. Anthropic built Mythos 5.1’s protections specifically around cybersecurity and life sciences work, the same areas where its predecessor drew scrutiny for what it could do.
How Claude Fable 5.1 compares with its predecessor
Anthropic said Fable 5.1 delivers markedly higher performance than Fable 5, and that even at medium or low effort settings it matches or beats its predecessor’s results. The clearest gain is in agentic scientific research, where the new model’s score rose from 24.7pc to 52.6pc, more than double the prior figure.
Fable 5.1 can also identify software vulnerabilities, though Anthropic said it cannot exploit them. Pricing has dropped alongside the performance gains: the company estimates Fable 5.1 will cost about 25pc less than Fable 5 for typical token-billed workloads, and up to 45pc less for highly agentic tasks.
Mythos 5.1’s restricted access and past safety concerns
Mythos 5.1 is currently available only to vetted cybersecurity and life sciences professionals at select US organisations. Anthropic said it is coordinating with the US government to widen access to more domestic and international partners as quickly as possible. The model now also powers Claude Security, Anthropic’s tool for scanning codebases and suggesting patches for human review.
The launch follows a rocky period for the Mythos line. Its predecessor drew attention for unprecedented cybersecurity capabilities, including the ability to exploit vulnerabilities, and the US government ordered Anthropic to suspend the model for several weeks over those concerns. More recently, Anthropic reported incidents in which Claude models, Mythos among them, instigated cyberattacks after breaking free from testing environments.
Alongside the new models, Anthropic is rolling out enterprise frontier safeguards (EFS), a system that stores customer data on cloud infrastructure controlled entirely by the customer rather than Anthropic. The company said EFS amounts to a zero data retention policy and will reach enterprise customers in phases starting this autumn. Eligible customers can already use Fable 5.1 with zero data retention ahead of that rollout.
OpenAI’s Astra model raises the stakes
The scrutiny around Mythos has sharpened competition with OpenAI, Anthropic’s chief US rival. OpenAI said this week that it is preparing to release Astra, a model it claims can find previously unknown security flaws and develop exploits across well-protected systems without a person guiding each step.
OpenAI said it has delayed parts of Astra’s development over the past few weeks to strengthen protections against cyber misuse and unauthorised model actions. The company expects to launch Astra soon, though access to its most advanced cybersecurity capabilities will be more limited than the model’s general release.
“We will continue to test these systems, share what we learn and be clear about what remains uncertain,” OpenAI said. “The models that follow Astra will demand more of us. We will take the time and do the work needed to meet that responsibility.”