Anthropic Unleashes Fable 5 and Mythos 5, Setting New AI Benchmarks with Strategic Rollout

Anthropic has officially released its highly anticipated Fable 5, touted as its most potent general-purpose AI model, alongside a specialized, less-guarded counterpart, Mythos 5. Fable 5 is declared to surpass all currently available general models, exhibiting state-of-the-art performance across numerous AI capability benchmarks. It shows exceptional prowess in software engineering, knowledge work, scientific research, and other critical areas, notably outperforming Opus 4.8 (69.2% agentic vs. Mythos 5’s 80%) and GPT 5.5 in agentic tasks and other metrics. Mythos 5, essentially Fable 5 without some of its built-in safeguards, is exclusively available to cyberdefenders and infrastructure providers, underscoring its dual-tier release strategy. Both models utilize the same API as Opus 4.7 and 4.8, with a notable breaking change that allows disabling the ‘thinking’ type, and offer a substantial context window of 300,000 tokens.

Demonstrations highlight Fable 5’s impressive generative capabilities, including the creation of an interactive 3D solar system (costing approximately $0.30), a lightweight 260-line React framework ($4.58), and a sophisticated 3D traffic simulator complete with configurable vehicles and semaphores. However, Fable 5’s integrated safety guardrails were evident, with the model reverting to Opus 4.8 for queries related to cybersecurity vulnerabilities or sensitive biology topics. The pricing structure positions Fable 5 at double the cost of Opus. Anthropic is offering Fable 5 to Pro, Max, and Team subscribers without extra charge until June 22nd, after which it will transition to a credit-based usage model. This aggressive market rollout is notably juxtaposed with Anthropic’s recent public advocacy for a collective pause or slowdown in frontier AI development among leading laboratories, a position that has not gone unnoticed by the developer community.