Skill Leap AI

Claude Fable 5 Is Here: Anthropic Just Took a Giant Lead

Claude unveiled two new models—Fable 5, now accessible via the Claude app and API, and Mythos 5, a restricted version with lifted cyber safeguards for select partners. The rollout adds a fourth tier to Anthropic’s lineup, sitting above Haiku, Sonnet, and Opus 4.8, and is positioned as the most capable LLM currently available. Benchmark data referenced in the video shows Fable 5 surpassing Opus 4.8 and outperforming competing models from Gemini and OpenAI across software engineering, knowledge work, and long‑running autonomous tasks. Users can select reasoning effort levels (high, extra, max), with higher settings boosting accuracy at the cost of token usage. Pricing is roughly double Opus—$10 per million input and $50 per million output tokens—and will transition to a pure usage model on June 22. The presenter highlights practical demos: a slideshow created in Claude Design, interactive simulations, and rapid revisions of visual assets, all showing marked quality gains over Opus. However, occasional policy blocks forced the system to fall back to Opus, and the model’s aggressive safeguards sometimes limit creative freedom. Mythos 5 remains behind a partner‑only wall due to concerns about enabling malicious cyber or bio‑research activities. For businesses, Fable 5 offers a performance leap for AI‑driven development and content generation, but cost and usage‑based billing demand careful budgeting. Anthropic’s tiered safety approach signals a broader industry trend of balancing power with responsible AI deployment.