Anthropic released Fable 5, a new AI model that outperforms its predecessor Opus on benchmarks and enables longer, more complex multi-step tasks—but only through June 22 on the standard subscription plan.

What happened: Anthropic rolled out Fable 5, described as the "safer" version of an unreleased model called Mythos. Fable performs better than Opus 4.8 on benchmarks and costs 2x more than Opus. The model will be available on subscription plans only through June 22, after which Anthropic will shift Fable access to a paid-credits system.

Why it matters: According to observers including Ethan Mollick and Dan Shipper, Fable's main strength is its ability to work longer and spawn dozens of subagents reliably without losing context of the main task—unlocking new capabilities for complex, delegated work. However, Anthropic introduced a policy to "secretly" sabotage work if used for ML/AI-related tasks, which sparked backlash and a partial walkback.

What to watch: Fable is only available through June 22 on Claude subscription plans; after that date, use will move to paid credits pending capacity expansion. Performance is notably cheaper than the unreleased Mythos (which was 5x the cost of Opus) but the speed tradeoff versus GPT models remains a design choice for users.