Claude Fable 5.2: benchmark results
Anthropic's Claude Fable 5.2 (claude-fable-5-2), released November 9, 2026, is available through Claude products and APIs, with a 1M-token context and 128K maximum output. Shares underlying model weights with restricted-access Claude Mythos 5.2; Fable applies additional cyber and biology safeguards. Default API effort is high; app and Claude Code default effort is medium. Native fallback/refusal settings remain separately attributed. Official release: https://www.anthropic.com/claude-fable-and-mythos-5-2 Provider: Anthropic. Released 2026-11-09. Access: API.
Aggregate Rating 2009 ± 9, rank #1 of 10870 rated models, from 337 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA CritPt v2 | 86.29 | Accuracy (%) | 100 |
| AA CritPt v2 - Hard Subset | 76.5 | Pass@1 (%) | 100 |
| AA FrontierCode 2.0 | 49.13 | Guardrail-gated Rubric Score (%) | 100 |
| AA GDP.pdf - Engineering - Criterion Pass Rate | 86.26 | Criterion Pass Rate (%) | 100 |
| AA HLE-Diamond | 80.2 | Accuracy (%) | 100 |
| AA HLE-Diamond - Knowledge | 74.8 | Accuracy (%) | 100 |
| AA HLE-Diamond - Reasoning | 85.6 | Accuracy (%) | 100 |
| AA Humanity's Last Exam | 74.28 | Accuracy (%) | 100 |
| AA MKA - Climb WSR | 56.56 | Weighted Satisfaction Rate (%) | 100 |
| AA MKA - Output File Type - pdf - Climb WSR | 60.71 | Weighted Satisfaction Rate (%) | 100 |
| AA MKA - Output File Type - spreadsheet - Climb WSR | 56.89 | Weighted Satisfaction Rate (%) | 100 |
| AA MKA - Output File Type - text - Regression WSR | 97.34 | Weighted Satisfaction Rate (%) | 100 |
Interactive version: theaggregate.ai/model?slug=claude-fable-5-2 · How It Works · Data refreshed daily, snapshot 2026-11-12.