RC RANDOM CHAOS

Anthropic ships Fable 5.1 and Mythos 5.1: cheaper tokens, vuln discovery, science gains

· via Hacker News

Original source

Claude Fable 5.1 and Claude Mythos 5.1

Hacker News →

Anthropic released two versions of the same underlying model: Fable 5.1, which is generally available, and Mythos 5.1, a gated release offered only through trusted-access programs with safeguards tuned for cybersecurity and life-sciences work. The launch answers customer complaints on three fronts. Pricing drops roughly 25% for typical token-billed workloads—and up to about 45% for heavily agentic use—by cutting the cost of cache reads. A new Enterprise Frontier Safeguards system will let enterprises keep data in infrastructure they control (equivalent to zero data retention) while still blocking adversarial use, rolling out in phases this fall. Reworked safeguards cut false positives in cybersecurity by 60%, and the model is now cleared to find software vulnerabilities, though not to build exploits for them.

On capability, Anthropic positions Fable 5.1 as a new bar for coding, knowledge work, and long-running tasks, with Low and Medium effort settings matching or beating the prior Fable 5 at lower cost. As a proof point, the firm Millennium says the model diagnosed the root cause of a rare internal crash that its engineers—and other models—had failed to explain for years.

The more striking claims are scientific. Mythos 5.1 designed protein binders with affinities ten times better than the top entries in Adaptyv Bio’s competitions, hitting a viable-binder rate near 50% across 12 targets versus a typical 10–15%. Fable 5.1 trained a neural network to build a higher-resolution elevation map covering a third of Venus from decades-old Magellan radar data, released under Creative Commons ahead of NASA and ESA missions. Mythos 5.1 also rewrote GPU kernels to speed up seven deep-learning models by up to 2.5x with identical outputs, cutting estimated compute costs 30–60%. On safety, Anthropic says both models were red-teamed for chemical, biological, and cyber risks; Mythos 5.1 exceeds its predecessor’s capabilities but still falls below the next tier in the Responsible Scaling Policy, so it ships with the same biology-access restrictions as Mythos 5.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.