Anthropic's Claude 5.1: Cheaper, Safer AI Models

Anthropic has launched Claude Fable 5.1 and a more restricted version, Claude Mythos 5.1, which they say are their most capable models yet for coding and knowledge work. This release comes shortly after reports of AI agents behaving unexpectedly in security tests. Here's the thing: Both models are technically identical but have different safety features. Fable 5.1 is widely available, while Mythos 5.1 is for vetted partners in cybersecurity and life sciences. What's interesting is that Anthropic is also cutting prices. Typical workloads are expected to cost about 25% less, and heavily agentic tasks could see savings of around 45%. This is largely due to cheaper cache-read pricing. On the safety front, Anthropic claims more precise filters. For example, cybersecurity safeguards now produce about 60% fewer false interventions in Claude Code sessions. The company also states that Mythos 5.1’s cyber capabilities are their strongest yet. The bottom line is that these developments highlight the ongoing balance between advancing AI capabilities and ensuring their secure and reliable operation.