<\/p>\n
After months of warning about the potential dangers of releasing Mythos, Anthropic is bringing its powerful model to the public, but with a few caveats. The new model is called Fable 5 and comes with restrictions in areas where Anthropic thinks it could be misused.<\/p>\n<\/div>\n
However, despite the restrictions, Fable 5 is claimed to be the most powerful AI model available not just from Anthropic, but from any AI provider at the moment.<\/p>\n<\/div>\n
Anthropic says Fable 5 outperforms every previous Claude model and has achieved state-of-the-art performance on nearly all tested AI benchmarks, including software engineering, vision, scientific reasoning, and knowledge work.<\/p>\n<\/div>\n
Anthropic also says that Fable 5 outperforms rival AI models from OpenAI and Google across a variety of benchmarks.<\/p>\n<\/div>\n
On SWE-Bench Pro, a benchmark that measures how well AI models can solve real-world software engineering problems, Anthropic says Fable 5 scored 80.3%, while GPT-5.5 (58.6%) and Gemini 3.1 Pro (54.2%) ranked behind even Claude Opus 4.8 (69.2%).<\/p>\n<\/div>\n
Similarly, on FrontierCode (Diamond), a benchmark that measures the real-world production quality of AI-generated code, Fable 5 scored 29.3%, while GPT-5.5 lagged behind at 5.7%.<\/p>\n<\/div>\n
Fable 5 also took the lead in the coveted Humanity’s Last Exam benchmark, which measures the limits of AI reasoning and the depth of scientific knowledge. Fable 5 came out on top with a score of 59% without tool use and 64.5% with tools, while GPT-5.5 <\/a>and Gemini 3.1 Pro<\/a> trailed with scores of 41.4% and 44.4%, respectively.<\/p>\n<\/div>\n Anthropic says that during early testing with Stripe, Fable 5 was able to compress months of software engineering work into days. The AI model is said to have completed a 50-million-line Ruby codebase migration in a single day that would otherwise have required a team of engineers working for more than two months.<\/p>\n<\/div>\n Anthropic says Fable 5 is also excellent at tasks that require vision capabilities. This means the model can extract information from scientific charts, understand complex diagrams, and even recreate a web application’s source code using screenshots alone.<\/p>\n<\/div>\n3) Fable 5 can compress months of coding into days<\/h2>\n
4) Rebuild apps from screenshots<\/h2>\n