LongCat 2.0: benchmark results
Meituan's open-weights 1.6T-A48B mixture-of-experts model with native 1M context via LongCat Sparse Attention. Provider: Meituan. Released 2026-07-05. Access: Open.
Unified ELO 1635 ± 1, rank #180 of 1553 rated models, from 49 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| WritingBench | 83.8 | Score (self-reported) | 98.1 |
| AA Omniscience - Software Engineering (SWE) - Go | 38.78 | Accuracy (%) | 92.5 |
| AA Omniscience - Software Engineering (SWE) - Dart | 32.65 | Accuracy (%) | 88.6 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 34 | Accuracy (%) | 88.1 |
| AA Omniscience - Software Engineering (SWE) - JavaScript | 45.45 | Accuracy (%) | 86.9 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 42.22 | Accuracy (%) | 86.9 |
| AA Omniscience - Software Engineering (SWE) - Swift | 52 | Accuracy (%) | 86.7 |
| AA Humanity's Last Exam | 33.69 | Accuracy (%) | 85.3 |
| AA Omniscience - Software Engineering (SWE) - Julia | 28 | Accuracy (%) | 85.2 |
| AA Omniscience - Software Engineering (SWE) - Python | 33 | Accuracy (%) | 84.3 |
| AA Omniscience - Software Engineering (SWE) - C | 53 | Accuracy (%) | 84.1 |
| AA Omniscience - Software Engineering (SWE) - R | 24 | Accuracy (%) | 83.4 |
Interactive version: theaggregate.ai/model?slug=longcat-2-0 · How It Works · Data refreshed daily, snapshot 2026-09-08.