Ling-2.6-1T: benchmark results
InclusionAI's open 1T-parameter Ling 2.6 MoE, the non-reasoning counterpart to Ring-2.6-1T. Provider: InclusionAI. Released 2026-04-23. Access: Open.
Aggregate Rating 1651 ± 20, rank #309 of 1656 rated models, from 89 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FinRiskAtlas - Evidence-Grounded Processing - Quantitative Reasoning | 86.21 | Numerical Accuracy (%) | 100 |
| FinRiskAtlas - FinRisk-Ask - Evidence-Request Alignment | 79.75 | ERA (%; DeepSeek-V4-Flash request-alignment judge) | 100 |
| FinRiskAtlas - Applied Review - Risk Classification | 75.68 | Accuracy (%) | 95.3 |
| AA Omniscience - Software Engineering (SWE) - HTML | 48 | Accuracy (%) | 85.8 |
| FinRiskAtlas - Domain Knowledge - FinTech & Security | 83.22 | Macro-Averaged Accuracy (%) | 84.4 |
| AA TAU-2 Bench | 89.77 | Accuracy (%) | 84.3 |
| Tau2-Bench Telecom | 89.77 | Success Rate (%) | 84.2 |
| AA Omniscience - Software Engineering (SWE) - JavaScript | 40.91 | Accuracy (%) | 83.4 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 36.67 | Accuracy (%) | 82.4 |
| FinRiskAtlas - FinRisk-Ask - Conditional Request Alignment | 82.58 | CRA (%; DeepSeek-V4-Flash request-alignment judge) | 82.3 |
| AA Omniscience - Software Engineering (SWE) - Dart | 28 | Accuracy (%) | 82.1 |
| AA Omniscience - Software Engineering (SWE) - Java | 23 | Accuracy (%) | 81.7 |
Interactive version: theaggregate.ai/model?slug=ling-2-6-1t · How It Works · Data refreshed daily, snapshot 2026-10-20.