Ring-flash-2.0 — benchmark results

InclusionAI's open 100B MoE thinking model (6.1B active), the RL-trained reasoning counterpart of Ling-flash-2.0 (September 2025). Provider: InclusionAI. Released 2025-09-19. Access: Open.

Unified ELO 1533 ± 69, rank #673 of 1776 rated models, from 38 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA AIME 202583.67Accuracy (%)80.4
AA LiveCodeBench62.75Pass@1 (%)68.7
CritPt0.3Accuracy (self-reported)65
AA MMLU-Pro79.27Accuracy (%)63.4
AA Humanity's Last Exam8.94Accuracy (%)58.3
SuiChat-CN68.7Standard F1 (self-reported)58.1
AA GPQA Diamond72.53Accuracy (%)56.9
AA CritPt0.28Accuracy (%)54.4
AA Omniscience - Science, Engineering & Mathematics26.1Accuracy (%)50.4
AA Omniscience - Software Engineering (SWE) - Kotlin18Accuracy (%)50
AA Omniscience - Software Engineering (SWE) - Julia12Accuracy (%)47.9
AA Omniscience - Software Engineering (SWE) - Swift36Accuracy (%)47.7

Interactive version: theaggregate.ai/model?slug=ring-flash-2-0 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.