Fable 5.1 handles extremely long contexts with ease — its one-million token window means it can hold entire codebases, lengthy legal documents, or book-length material in a single conversation without losing the thread. It works through text-only inputs and outputs, keeping its focus narrow but deep. Details about its specific reasoning style or capability trade-offs beyond context handling are limited from available data.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| LCR | 85.3 | accuracy | 29d ago |
| LiveCodeBench | 90.5 | accuracy | 29d ago |
| GPQA Diamond | 93.7 | accuracy | 29d ago |
| MMLU-Pro | 92.4 | accuracy | 29d ago |
| SWE-Bench | 81.2 | accuracy | 29d ago |
| Humanity's Last Exam | 60.9 | accuracy | 29d ago |
| SciCode | 63.1 | accuracy | 29d ago |