Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model

TL;DR

This paper introduces Ring-1T, a trillion-parameter reasoning model with novel training innovations, achieving state-of-the-art benchmarks and enabling open access to large-scale reasoning capabilities.

Contribution

It presents three new methods—IcePop, C3PO++, and ASystem—for training trillion-scale models efficiently and reliably, and releases the first open-source trillion-parameter reasoning model.

Findings

Achieved high benchmark scores across multiple reasoning tasks.

Demonstrated stable training of trillion-parameter models.

Enabled open access to a state-of-the-art reasoning model.

Abstract

We present Ring-1T, the first open-source, state-of-the-art thinking model with a trillion-scale parameter. It features 1 trillion total parameters and activates approximately 50 billion per token. Training such models at a trillion-parameter scale introduces unprecedented challenges, including train-inference misalignment, inefficiencies in rollout processing, and bottlenecks in the RL system. To address these, we pioneer three interconnected innovations: (1) IcePop stabilizes RL training via token-level discrepancy masking and clipping, resolving instability from training-inference mismatches; (2) C3PO++ improves resource utilization for long rollouts under a token budget by dynamically partitioning them, thereby obtaining high time efficiency; and (3) ASystem, a high-performance RL framework designed to overcome the systemic bottlenecks that impede trillion-parameter model training. Ring-1T delivers breakthrough results across critical benchmarks: 93.4 on AIME-2025, 86.72 on HMMT-2025, 2088 on CodeForces, and 55.94 on ARC-AGI-1. Notably, it attains a silver medal-level result on the IMO-2025, underscoring its exceptional reasoning capabilities. By releasing the complete 1T parameter MoE model to the community, we provide the research community with direct access to cutting-edge reasoning capabilities. This contribution marks a significant milestone in democratizing large-scale reasoning intelligence and establishes a new baseline for open-source model performance.