Falcon-H1R-7B: benchmark results
TII's 7B hybrid Transformer-Mamba2 reasoning model with a 256K context (January 2026). Provider: TII. Released 2026-01-06. Access: Open.
Unified ELO 1524 ± 1, rank #575 of 1392 rated models, from 42 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CritPt | 30 | Accuracy (self-reported) | 80.8 |
| AA LiveCodeBench | 72.38 | Pass@1 (%) | 77.5 |
| AA AIME 2025 | 80 | Accuracy (%) | 73.3 |
| AA IFBench | 54.42 | Accuracy (%) | 65.3 |
| MathArena - SMT 2025 | 85.85 | Accuracy (%) | 60.7 |
| AA Humanity's Last Exam | 10.98 | Accuracy (%) | 57.1 |
| MathArena - HMMT Feb 2025 | 84.17 | Accuracy (%) | 56.8 |
| MathArena - AIME 2025 | 86.67 | Accuracy (%) | 54.9 |
| AA CritPt | 0.29 | Accuracy (%) | 53.2 |
| AA Omniscience - Software Engineering (SWE) - Swift | 32 | Accuracy (%) | 51.2 |
| AA Omniscience - Software Engineering (SWE) - JavaScript | 27.27 | Accuracy (%) | 48.3 |
| AA Omniscience - Software Engineering (SWE) - Julia | 8 | Accuracy (%) | 46.1 |
Interactive version: theaggregate.ai/model?slug=falcon-h1r-7b · How It Works · Data refreshed daily, snapshot 2026-09-05.