Claude Fable 5.2: benchmarks, price and rank

Its headline results

The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.

• Remote Labor Index41.3%

  Real freelance projects that people were paid to do, from design and video to architecture, games and data work. The model gets the same brief and files the freelancer got and has to hand back finished work.

  1st of 17. Independent result from Epoch AI.

• APEX-Agents80.1%

  Long tasks written by investment banking analysts, management consultants and corporate lawyers from their own work, done across files and apps in a realistic work setup.

  3rd of 56. Independent result from Epoch AI.

• Terminal-Bench 4.067.6%

  Hard jobs done in a real computer terminal: building and fixing software, setting up systems, and science and engineering problems.

  2nd of 17. The maker's own number, from Anthropic, via LLM Stats.

• OSWorld 2.154.6%

  The newer version of OSWorld, with longer desktop tasks that are graded at several checkpoints along the way.

  1st of 6. The maker's own number, from Anthropic, via LLM Stats.

• SWE-Bench Pro93%

  Long changes to real codebases in several languages, often across many files, including code from private and copyleft projects that models are unlikely to have seen.

  1st of 50. The maker's own number, from Anthropic, via LLM Stats.

• FrontierSWE69.9%

  Open-ended technical projects that take hours to tens of hours, from making systems faster to building large programs and applied machine learning research. Each run gets a 20-hour budget.

  1st of 21. Independent result from Epoch AI.

• Humanity's Last Exam73.2%

  About 2,500 questions written by experts in more than a hundred fields to be as hard as they could make them, many needing a diagram or image.

  1st of 74. The maker's own number, from Anthropic, via LLM Stats.

• ARC-AGI-379.3%

  Interactive puzzle games the model has never seen, with no instructions. It has to explore, work out the goal and the rules, and win.

  2nd of 6. The maker's own number, from Anthropic, via LLM Stats.

• GDPval-AA1,831

  Work products from dozens of occupations, the kind of documents, spreadsheets and plans professionals hand in, made by the model and compared head to head with other models' work.

  4th of 5. The maker's own number, from Anthropic, via LLM Stats.

• CursorBench60%

  Ambiguous coding tasks that span many files, taken from real sessions in the Cursor code editor.

  2nd of 20. Independent result from Epoch AI.